This is the mail archive of the gcc-patches@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: libiberty: hashtab allocation functions with an extra argument


On Tue, Jan 14, 2003 at 04:34:12PM -0800, Stan Shebs wrote:
> Daniel Jacobowitz wrote:
> 
> >On Tue, Jan 14, 2003 at 07:04:24PM -0500, DJ Delorie wrote:
> >
> >>>Is this not feasible?
> >>>
> >>I don't know.  Why can't you use the regular malloc/free?
> >>
> >
> >Mmalloc supports saving and restoring allocated regions from a file,
> >via mmap.  Theoretically, we use it to write loaded symbol data to a
> >file and read it in again; quicker and more memory-friendly than
> >rebuilding it each session.
> >
> >I don't know if it works in current CVS GDB; but I'm pretty sure
> >Apple's using it and they keep talking about contributing the rest of
> >the fixes for it back to us.
> >
> Our PFE implementation uses mmap but not mmalloc.  As it happens,
> PFE seems likely to go away in favor of PCH, which is more robust
> and (I hope I'm not stealing Geoff's thunder :-) ) appears to be
> faster as well, at least in the early heats.

Hey, Stan, while you're out there - could you confirm or deny the thing
about caching objfiles from GDB using mmalloc()?  Is that still alive?

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer


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