This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
performance regressions, profiling _of_ gcc
- From: Jan Van Dijk <janvandijkinjapan at yahoo dot co dot jp>
- To: gcc at gcc dot gnu dot org
- Date: Wed, 9 Apr 2003 10:55:01 +0000
- Subject: performance regressions, profiling _of_ gcc
- Organization: TU/Eindhoven
- Reply-to: jan at etpmod dot phys dot tue dot nl
Hi,
In view of the recent compile time regressions in gcc versions I wanted to do
some (detailed) profiling of gcc (mostly g++, I am particularly interested in
the C++ front end).
To this end I need more detailed information than -ftime-report gives me.
I did some googling, but found no satisfactory links. For example, there was
no response to message http://gcc.gnu.org/ml/gcc/1999-11n/msg00333.html.
The configuration docs ( http://gcc.gnu.org/install/configure.html ) briefly
mention the --enable-coverage option. I read these and decided to configure
gcc (HEAD) with
--enable-coverage=opt --disable-checking
and bootstrapped the compiler, but I am clueless how to proceed. I was
assuming that running the new compiler on some test file would result in
files suitable for analysis _of gcc_ with gcov/gprof, but nothing seems to
appear.
Was I mistaken? Are there any other options? Is it possible to tell configure
to simply add the extra cflag `-pg' to stages 2 and 3 of the compiler build?
Should an external tool, like valgrind rather be used for performance
analysis?
Pointers to (well-hidden :-) docs about profiling gcc are appreciated? If no
such docs exist, brief instructions would be welcome, and I will try to
contribute some documentation myself.
Thanks in advance,
Jan van Dijk
--
Keio-Tsuushin Residence
Jan van Dijk, Room 210
2 Chome 19-30, Mita Minato-ku
108 Tokyo, Japan
jan at etpmod dot phys dot tue dot nl
tel: +81 3 5476 9461 (home)
__________________________________________________
Do You Yahoo!?
Yahoo! BB is Broadband by Yahoo! http://bb.yahoo.co.jp/