This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Haney's real matrix test regression
- To: kenner at vlsi dot ultra dot nyu dot edu
- Subject: Haney's real matrix test regression
- From: Paolo Carlini <pcarlini at unitus dot it>
- Date: Wed, 24 Oct 2001 13:37:25 +0200
- CC: gcc at gcc dot gnu dot org
- Organization: Universita' della Tuscia
- Reply-To: pcarlini at unitus dot it
Hi,
today, having updated my local 3.1 tree, I'm seeing a bad performance regression
(2x running time) in the C++ version of the real matrix test of the Haney suite,
wrt the perfect results of yesterday (i686-pc-linux-gnu, by the way).
I strongly suspect the MEM tracking checkins... It is possible?
In that case, I cannot understand if this is expected 'til the end of that
series of checkins or not: should I start right now pinning down the actual
assembly section responsible for the slow down or wait a few days more?
Opinions?
Thanks,
Paolo Carlini