This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Large performance regressions from 2.95.2/3 to HEAD
- To: dan at cgsoftware dot com
- Subject: Re: Large performance regressions from 2.95.2/3 to HEAD
- From: "Martin v. Loewis" <martin at loewis dot home dot cs dot tu-berlin dot de>
- Date: Fri, 30 Jun 2000 08:36:29 +0200
- CC: gcc at gcc dot gnu dot org
- References: <Pine.LNX.4.21.0006291224270.8609-100000@propylaea.anduin.com>
> Anyone have any idea what's going on?
From the bird's perspective, no. There are ways to find out the
problems quite easily: pick one of the test cases that has degraded
performance, generate assembler code with 2.95 and with the mainline,
and compare them. Try to ignore minor difference, instead detect the
major problems - profiling is your friend.
When you find anything on that level, don't hesitate to post your
results. The next step then is to analyse where these problems come
from - that, of course, requires a detailed understanding what the
problems *are*, at first.
Regards,
Martin