This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: 50% memory, 40% compile time regression the last two days
- From: Mark Mitchell <mark at codesourcery dot com>
- To: Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>
- Cc: gcc at gcc dot gnu dot org
- Date: Fri, 16 Jul 2004 13:34:02 -0700
- Subject: Re: 50% memory, 40% compile time regression the last two days
- Organization: CodeSourcery, LLC
- References: <Pine.LNX.4.53.0407161017480.2199@bellatrix.tat.physik.uni-tuebingen.de>
Richard Guenther wrote:
Hi!
This time I waited 2 days before reporting, just in case it was noise
(or promptly fixed) again. For the leafified tramp3d-v3 testcase on
ia64 we now need 6.2min for compilation where two days ago we needed
only 3.9min. At the same time memory requirements sky-rocketed from
about 1Gb to 1.4Gb! (see
http://www.tat.physik.uni-tuebingen.de/~rguenth/gcc/monitor-summary.html)
Patches causing this may be up to
My patches are certainly significant enough to be causing a problem.
I'd be a bit surprised, since, for the most part, they've been
*reducing* memory usage -- for example, I'd be surprised if the bits to
do lazy allocation of assignment operators could result in significantly
higher memory usage. But it's certainly possible that there's something
boneheaded in there!
Would you mind isolating the particular patch that caused the problem?
That would help a lot. Also, once you find the patch, if you can post
preprocessed source code and command-line options that trigger the
problem? Please copy me directly on mail that you post about this; I
sometimes get behind on the mailing list.
Thanks,
--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com