Re: memory allocations.
- From: Bill Gribble <grib linuxdevel com>
- To: Havoc Pennington <hp redhat com>
- Cc: Sander Vesik <sv117949 ireland sun com>, Ulrich Drepper <drepper redhat com>, Alan Cox <alan redhat com>, Chris Chabot <chabotc reviewboard com>, iain <iain ximian com>, gnome-hackers gnome org
- Subject: Re: memory allocations.
- Date: 01 Mar 2002 15:55:06 -0600
On Fri, 2002-03-01 at 14:03, Havoc Pennington wrote:
> Sander Vesik <sv117949 ireland sun com> writes:
> > Well, as I see it the problem with optimising malloc for startup times is
> > that according to my profiling, at most 1% time of total is spent in
> > malloc during startup of say gedit, while in real use - I opened a 300KB+
> > text file and page-downed to the end - the use grew to 3% of time.
>
> If this is gedit2 could you post the top of the profile? (Does it have
> call graph info?)
Havoc et al, I'm a little curious why the fascination with memory
allocation behavior. As Sander points out, for gedit the total time
spent in malloc is less than 3%. I would guess this is true across the
board for Gnome apps.. I know it has been for my apps that I've profiled
(including Gnucash). There's really not that much room for improvement;
even if you decrease total time spent in malloc and friends by 25% you
will only improve overall performance by a fraction of a percent.
Seems like reducing the number of redundant font loads, repeated network
fetches of the same piece of info, constant path walking to find a
library, etc., provide much more "low-hanging fruit". I agree that
chunking related allocations can help with fragmentation concerns, but
that seems not to be the focus of this thread.
Am I missing something?
Thanks,
Bill Gribble
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]