[olpc-software] graceful handling of out-of-memory conditions

[Jim Gettys]

On Mon, 2006-03-27 at 12:02 -0500, Alan Cox wrote:
> On Mon, Mar 27, 2006 at 11:19:54AM -0500, Jim Gettys wrote:
> > those applications are not even appropriate for the target audience.
> > I'm certainly going to talk about this at GUADEC.  I hope observations
> > about our memory wastage will be a wakeup call to the community.
> 
> With certain notable exceptions (evolution, where even security holes are
> being ignored nowdays) the core GNOME has been getting better over time and
> some of the really dumb stuff (but not all of it) has gone away. The lower
> level stuff like gtk is pretty tight although the string handling stuff seems
> to have the odd pathalogical disagreement with the glibc malloc.

I'm seriously considering abandoning evolution at this point.  For what,
is the question.

Yeah, our experiences on GPE/Maemo using GTK were positive.  Owen, et.
al. do fine work.

> 
> > The paging system should be recovering unused pages of text when more
> > heap/stack is needed; we certainly observed this behavior on the iPAQ.
> > But there comes a minimum point when you end up paging text too much out
> > of flash (which wastes power and hurts speed).
> 
> We recover roughly on the basis of "Least regularly used" and code pages are
> scavenged as well as data.  If you swap is flash you probably want to turn
> down some of the swap prefetch/block sizes as those are designed to try
> and beat sane performance out of modern hard disks.

Interesting point; for us, pretty small block sizes for code would make
lots of sense.  And prefetch makes no sense at all; access time is way
lower than a disk drive.
                              - Jim


-- 
Jim Gettys
One Laptop Per Child