This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Re: Doubling of 3.1's Running Time for i686-pc-linux-gnu
- To: oldham at codesourcery dot com
- Subject: Re: Doubling of 3.1's Running Time for i686-pc-linux-gnu
- From: Andreas Jaeger <aj at suse dot de>
- Date: Thu, 26 Jul 2001 07:18:57 +0200
- Cc: gcc-bugs at gcc dot gnu dot org
- References: <200107251732.KAA29388@oz.codesourcery.com>
Jeffrey Oldham <oldham@codesourcery.com> writes:
> My nightly gcc 3.1 builds and tests now require about double the
> amount of time compared with those on 19Jul or before.
Jeffrey,
what's the result with this patch applied (already in CVS)?
Thu Jul 26 00:19:30 CEST 2001 Jan Hubicka <jh@suse.cz>
* predict.c (estimate_probability): Avoid duplicated predictions.
Honza mentions:
>
> I've run into problem with memory explosion on compiling insn_extract
> with enable_checking. This is caused by the postdominators problems
> discussed earlier making predict.c to predict each NORETURN patch
> as many times as there are noreturns. This is really numbuers with
> enable_checking making gcc to require rought 300MB memory.
I also noticed some failures in the testsuite which lead to longer
runtimes. Can you compare the runtime with the number of failures you
get?
Andreas
--
Andreas Jaeger
SuSE Labs aj@suse.de
private aj@arthur.inka.de
http://www.suse.de/~aj