This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: sjlj CFG improvement
- To: law at redhat dot com
- Subject: Re: sjlj CFG improvement
- From: Marc Espie <espie at quatramaran dot ens dot fr>
- Date: Fri, 8 Dec 2000 04:33:41 +0100
- Cc: gcc at gcc dot gnu dot org
- Organization: Ecole Normale Superieure (quatramaran)
In article <19083.976233430@upchuck> you write:
>
> In message <XFMail.20001207130736.ak03@gte.com>you write:
> > Should this patch be backported for use by the upcoming GCC-2.95.3 release?
>I would recommend against it. Typically that kind of release would be to
>fix code generation bugs, aborts and the like. Stuff that is strictly
>a compile time improvement (even one as large as this) wouldn't qualify for
>inclusion into a minor release.
Consider that AGAIN. Those sjlj problems are making it almost impossible
to build a few pieces of software on sjlj machines.
This is really a critical issue. It makes a HUGE size difference
on some very important programs, like qt2 or kde2.
Now, I hope to make dwarf2 exception unwind work under OpenBSD when I get
time to do it (at least I know that these files should take much less memory
to compile if things were normal), but in the mean time,
some qt2 files require 250Mb to compile under -O2. Some kde2 files don't
compile AT ALL. I've had the box I'm running on go all the way up to
400Mb of virtual memory space before giving up, after over an hour
of compilation.
This is a PIII 700, btw... not exactly a slow machine.