This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Pessimization from reduce-all-givs and move-all-movables
- From: Jan Hubicka <jh at suse dot cz>
- To: Toon Moene <toon at moene dot indiv dot nluug dot nl>
- Cc: Jerry Quinn <jlquinn at us dot ibm dot com>, gcc at gcc dot gnu dot org
- Date: Wed, 20 Nov 2002 21:54:51 +0100
- Subject: Re: Pessimization from reduce-all-givs and move-all-movables
- References: <E18EO1A-00029F-00@tiamat> <3DDBED2A.20909@moene.indiv.nluug.nl>
> Jerry Quinn wrote:
>
> >Without -fmove-all-movables -freduce-all-givs, performance improves to
> >1048 Kpos/sec, a 7% difference.
>
> This wouldn't surprise me. Between when I introduced those flags (in
> 1997) and now we gathered quite some information on them. It turns out
> they're mostly useful for Fortran code (subroutines) with rank-n dummy
> arguments where n>1. Because this situation is so common in Fortran
> code, these options are enabled by default when invoking g77.
>
> For all other types of code it's a loss, especially on targets with very
> few integer registers, like the Intel x86 line and derivatives.
>
> Hope this helps [I can make the above reasoning more precise if you're
> interested],
Perhaps it would make sense to add this into manual instead of call for
testcases...
Honza
>
> --
> Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
> Saturnushof 14, 3738 XG Maartensdijk, The Netherlands
> Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
> Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)