FW: Squeak on iPaq..success!

Peter Crowther Peter.Crowther at melandra.com
Fri Jan 5 16:22:07 UTC 2001


Cue red face.  One of these days we'll sort out the reply-to on the list
(I'm on Randal's side on this one :-).

		- Peter

-----Original Message-----
From: Peter Crowther 
Sent: 02 January 2001 11:25
To: 'tim at sumeru.stanford.edu'
Subject: RE: Squeak on iPaq..success!


> From: Tim Rowledge [mailto:tim at sumeru.stanford.edu]
> I've claimed for years that the key determinant of speed for Smalltalk
> is memory bandwidth.

As with most OO systems, I think.  Smalltalk would probably have worse
locality of reference than most as the methods are spread pretty evenly
through object memory, which is probably why it plays merry hell with the
paging algorithms on most OSs.

I also suspect (but can't prove) that Smalltalk won't make as much use of
the fast burst read facilities of most modern systems, as there'll be fewer
sequential accesses to anything except the dispatch loop --- most of which
should be in the i-cache.  Anyone got any better idea on this?

> What I really want is a nice simple machine with a gigaHertz
> clock and memory that can keep up :-)

BBC Micro on speed? Anyone fancy a 6502 overclocked to a gigahertz? :-)

		- Peter





More information about the Squeak-dev mailing list