Carl,

> Is the VM with slider non GAPI friendly, since my Maestro seems not to
> support GAPI (or maybe has an incompatible GAPI)?

  The GAPI in SqueakWithMemorySlider.zip VM worked fine for Russell
(Allen)'s iPAQ on which the old VM didn't work properly.  It appeared
that the PocketPC OS on 38xx series has a bug in GAPI, but the newer
development environment I recently started using seems to include the
workaround for it.

  I think it is worth to try the new VM on yours and there is fair
chance it works.

> How does Win CE run so nice and snappy.  Is it just a matter of it being
> written in C/C++?  Is Squeak really a viable platform for this class of
> machine?

  I would guess that memory access pattern is pretty different in
Squeak and the lack of the secondary cache and slow memory bus hurt
the performance.  But this is my guess.

  While Morphic runs slowly, using MVC or other simple GUI frameworks
is definitely viable.  John Maloney and I (err, mainly by John, of
course) implemented such small GUI framework and that runs just fast
enough on 206MHz iPAQs, or even on 60MHz Hitachi SH-3.

> I can't set to full screen mode (or at least it seems to do nothing when I
> do).  Is this a GAPI related feature?

  In my implementation, yes it is.  But, Mike Mueller (it is actually
u-umlaut, I guess) sent me a patch that let us to switch to full
screen mode without GAPI turned on.

> >   (On XScale 400MHz, it is about 1900 msecs.)
> 
> On my 206Mhz SA machine, 7964 msecs.

  That sounds right.  I bet the GAPI will make it 5-6 times faster and
make it faster than XScale machine.

> >   (On XScale 400MHz, it is about 1300 msecs.)
> 
> On my 206MHz SA machine, 828 msecs.  Yes, that's 828.  Funny.

  This parallels with the above result, but somewhat bigger difference
I imagined.  (What's funny about 828?)

  Maybe the slower memory bandwidth and the write-through cache
problem make XScale look that slower?

-- Yoshiki