[Sugar-devel] [PATCH sugar 01/20] Move the sugar-session code inside jarabe

[James Cameron]

Thanks.  Good method, good data, no significant regression on your
laptop.

Yes, a test on an XO would be interesting.

Yes, buildbot supervised timing would be interesting.

Yes, activity startup should be a priority too.

Nine seconds seems a lot.  I wonder what is consuming that time?

On Tue, Dec 11, 2012 at 11:49:24PM +0100, Daniel Narvaez wrote:
> Some initial data when running on my laptop:
> 
> * I dropped kernel caches  with "echo 3 > /proc/sys/vm/drop_caches"
> * The start time has been taken as early as possible, after importing
> sugar3.logger and logging
> * The end time is at the first map event on the HomeWindow (seems to
> correspond well with the perceived UI appeareance).
> 
> On the old code it takes ~ 9 seconds, on thew new one ~ 9.1. So a ~ 1%
> regression which I don't think is very relevant, it's small and it's
> probably going to be spent waiting for the UI to be fully active
> anyway.
> 
> I'll see if I can repeat the same on the XO one of these days, of
> course things might be different there.
> 
> Not exactly on topic but I feel this could be really improved without
> making the code unreadable if we had buildbot timing the UI test and
> reporting regressions/improvements (though perhaps it would be better
> to focus on activiities startup). Well, maybe when I'm done with
> automated testing... :)
> 
> On 10 December 2012 23:00, James Cameron <quozl at laptop.org> wrote:
> > No, I do not recall sorry.
> >
> > Thanks, I acknowledge you aren't measuring yet.
> >
> > There is a potential for startup performance impact with these
> > changes.  There is a risk of regression.
> >
> > On Mon, Dec 10, 2012 at 10:39:46PM +0100, Daniel Narvaez wrote:
> >> Do you remember what/how it was measured maybe?
> >>
> >> Even if there are not idles we are constructing the home as very first
> >> thing and then after window manager loading we are doing all the rest.
> >> I suspect in theory that's all you can gain but... of course I could
> >> be proved wrong by measuring :)
> >>
> >> On 10 December 2012 21:27, James Cameron <quozl at laptop.org> wrote:
> >> > Have you any Sugar startup time measurements to share; before and
> >> > after this patch set?
> >> >
> >> > I seem to recall the movement of startup processing into idle did have
> >> > positive effects at the time, but since then much has changed in the
> >> > underlying software stack, and so new tests are needed.
> >> >
> >> > --
> >> > James Cameron
> >> > http://quozl.linux.org.au/
> >>
> >>
> >>
> >> --
> >> Daniel Narvaez
> >
> > --
> > James Cameron
> > http://quozl.linux.org.au/
> 
> 
> 
> -- 
> Daniel Narvaez

-- 
James Cameron
http://quozl.linux.org.au/