This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Various GCC compared (2.95.3 - 3.5-20040523) III
- From: Rene Rebe <rene at rocklinux-consulting dot de>
- To: gcc at gcc dot gnu dot org
- Date: Thu, 03 Jun 2004 04:14:30 +0200 (CEST)
- Subject: Various GCC compared (2.95.3 - 3.5-20040523) III
Hi all,
My next "preleminary" results are now available for an athlon-xp:
http://rocklinux-consulting.de/new-rt.png
http://rocklinux-consulting.de/new-bt.png
(up-to-date g3 results will follow today after a sleep cycle and will
be named new-??-g3.png when available)
Notes:
* -athlon results are built with "-O2 -mcpu=athlon-xp (or athlon or
i686 or -xK respectively - depending on the compilers support)"
The other results are built with "-O2" only for reference
* There are some up and downs in the runtime. New versions do not always
produce faster code.
* my gcc-3.5-20040523 _has checking_ disabled
there clearly is a compile time speed regreesion for all tests
* The Botan (C++) bench runtime for 3.5 is not very low - but just
not yet present since g++-3.5-20040523 ICEs. Maybe this is due to
incorrect template code I used -fpermissive to work around - I'll
try to fix the code properly and see if it does not ICE anymore.
Some notes:
* all run GNU/Linux - Athlon-XP 2500
* all gcc builds are bootstrap builds, that is no profiledbootstrap
(if time permits I will add seperate results with profiledbootsrap)
* test description can be found at:
http://www.rocklinux-consulting.de/oss/openbench/tests.html
Tests I will add later:
ogg/vorbis: compression decompression
I still search for a more C++ testsuits I could incooperate (xpdf was
proposed).
Any comments welcome!
Sincerely yours,
René Rebe
--
René Rebe - Europe/Germany/Berlin
rene@rocklinux.org rene@rocklinux-consulting.de
http://www.rocklinux.org http://www.rocklinux-consulting.de