[Sugar-devel] Physics-26 on Sugar 0.104 on XO-1

[James Cameron]

On Thu, May 14, 2015 at 03:01:24AM +0100, Gary Martin wrote:
> Hi James,
> 
> > On 13 May 2015, at 23:13, James Cameron <quozl at laptop.org> wrote:
> > 
> > On Wed, May 13, 2015 at 06:51:02AM -0400, Adam Holt wrote:
> >> A [...] XO-1 has issues with Physics on 13.2.4 -- I don't have
> >> details yet (I hope [n] can report more what happened) but has
> >> anybody else had recent problems with the Physics activity in
> >> general?
> > 
> > OLPC OS 13.2.4 has Physics-26 on Sugar 0.104.  Physics was not in any
> > previous XO-1 release.
> > 
> > For me, it tests out fine.
> > For some people, it may seem slow.
> > 
> > On an XO-1 the CPU is pushed very hard by Physics.  This will cause:
> > 
> > 1.  every other activity to perform badly, if Physics is not
> > stopped before moving to another activity; so make sure you stop
> > Physics carefully,
> 
> It’s been a while since I was working on the activity code, but I
> did spend quite some time detecting and debugging when Physics was
> not the activity in current focus, and minimise its cpu usage (stops
> all physics calculations and any buffer redrawing). It does still
> continue to consume a little cpu (pygame run loop), along with
> perhaps memory it was using. I was certainly testing/developing on
> XO-1 as the primary use case, so if it is really making other
> activities perform badly while it is in the background then there
> may be some environment regression related to events being
> passed/triggered (it watches _focus_event and _window_event to help
> spot if it is in_focus).

Thanks Gary.

I've tested, it still works; Physics does relax from 100% demand down
to 11% on

- switch to journal,

- switch to another activity,

- switch to neighbourhood view, friends view, or home view, and;

- switch to virtual console,

However, it does not release memory, and continues at 25%, and grows
ever so slowly.  Memory is so short on XO-1 at 256 MB, and now that
we're using swap on compressed jffs2 a memory shortage is converted
into CPU demand which is charged to the next process.

So I think the best recommendation is to stop the activity if there
are system performance concerns.  ;-)

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