Yoshiki.Ohshima at acm.org wrote:

>   I also see this slower sends/sec value on my machines.  I'm feeling
> that the sends/sec number has slightly more correlation with the real
> situation, so it is kind of understandable that Squeak on PXA250 is
> slower than StringARM 266MHz(? Can you clock it up to this much?).
Yup. You can buy SA110 boards for Acorn RPCs clocked up to 300MHz. I
would be moderately surprised if the heat at that speed wasn't a problem
in tight spaces like a PDA though.  An SA 110 at ~200MHz is ambient temp
pretty much but by 300 it starts to warm up quite a bit.

bytecodes/sec is a trivial test of the cpu/cache where as sends/sec
exercises the main memory a lot on a mchine with only 32kb cache! This
is likely most of the reason that sends/sec figures are very close
between my wet-string memory bus Acorn and faster clocked cpus with
not-very-much-faster memory buses. Final figures are not in yet but the
new 600MHz XScale 80321 (or some similar number) Iyonix RiscOS
(www.iyonix.com) machine seems to be 3-4 times faster for Squeak. 200MHz
DDR ram makes up quite a bit for a tiny cache. Donations to help me
afford one are welcome.

Soon we should be able to have ARM1120's at 1.2GHz though. :-) At least
that's what Samsung people tell me. I'm having lunch with an ARM staffer
next week, so maybe there will be more good news after that?

tim
-- 
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
Know Thy User.