This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Recent RSS growth for our NWP code.
- To: Mark Mitchell <mark at markmitchell dot com>, egcs at egcs dot cygnus dot com
- Subject: Recent RSS growth for our NWP code.
- From: Toon Moene <toon at moene dot indiv dot nluug dot nl>
- Date: Sat, 27 Feb 1999 15:07:34 +0100
- Organization: Moene Computational Physics, Maartensdijk, The Netherlands
[ Sorry for the clumsy reply - my ISP decided to spread the wrong info
on the mail route to moene.indiv.nluug.nl again [yes, I've notified
them], so I'm reading this with lynx on the egcs web archives -
Netscape dies a horrible death trying to read in 264K of archive ]
Mark,
You wrote:
> The most likely cause for the stack-frame expansion is this change, by
> yours truly.
> If -fno-strict-aliasing makes the expansion go away, that will confirm
> this hypothesis.
The interesting thing is that I indeed thought about your alias-set
patches as a possible culprit, but Fortran doesn't allow array
temporaries (at least not the language that g77 supports), so how can
this be a problem ? The only reuse of stack space (in the Mbyte size
category) is *between* subsequent SUBROUTINEs ...
Nevertheless, I'll try your suggestion and report back on it.
Just now I'm testing gcc version egcs-2.91.62 19990224 (egcs-1.1.2
pre-release-2) and thinking about a patch to have a flag to turn off the
no-alias assumption on struct/varying address vs. scalar in alias.c
Cheers,
--
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
g77 Support: fortran@gnu.org; egcs: egcs-bugs@cygnus.com