This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: New loop unroller broken?
- From: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- To: Mircea Namolaru <NAMOLARU at il dot ibm dot com>
- Cc: Dale Johannesen <dalej at apple dot com>,David Edelsohn <dje at watson dot ibm dot com>, gcc at gcc dot gnu dot org,Andrew Pinski <pinskia at physics dot uc dot edu>,Ulrich Weigand <weigand at i1 dot informatik dot uni-erlangen dot de>
- Date: Thu, 29 Jan 2004 13:19:50 +0100
- Subject: Re: New loop unroller broken?
- References: <OFFC3CC7AD.F3A9B29E-ON42256E2A.003A892D-42256E2A.004159B0@il.ibm.com>
Hello,
> We modified our patch following Zdenek's comments (thanks!).
> This patch gains more than 4% improvement overall CFSPEC2000 on Power4,
> with 3 benchmarks showing around 10% improvement (wupwise, swim and art).
btw. could you please measure the difference between
-fbranch-count-reg -funroll-loops
and
-fno-branch-count-reg -funroll-loops
with your patch? If it showed up to be small enough (and it might be the
case -- bct optimization as done in gcc does not bring much profit
anyway), it would probably be easier and cleaner to just switch off the
bct when unrolling is done.
Zdenek