Embedded Squeak, was Re: Introduction

Tim Rowledge tim at sumeru.stanford.edu
Fri Mar 8 18:56:00 UTC 2002


"Les Tyrrell" <tyrrell at canis.uiuc.edu> is claimed by the authorities to have written:


> Tim Rowledge has apparently done ports of Squeak to "bare" metal ( from what
> I've been able to gather, anyway ), perhaps several times.
Quite so. Fun,fun,fun. 'specially when somebody pays for it!

[snip]
> I think there are other examples.  I think that the memory constraints that we
> both are looking at are a bit tight, but not totally out of reach.  Since I've
> been working under the assumption that "embedded" devices are basically going
> to have a lot more RAM by the time I need to worry about this, the main
> stumbling block that I see is that the VM's ObjectMemory & Interpreter may
> need some modification in order to take advantage of sections of memory that
> are ROM vs. those that are RAM.  I'm not a VM guru by any means, so others can
> comment on that.
You're pretty much dead-on. Making a part-rom prat-ram image with the
current object memory would be tricky though probably not impossible.
Experience suggests that an Object Table can really earn its keep for
this purpose.

Another possibility is using flash (or other sorta-writable) memory.
It's almost like ram but you can write to it at a cost. That would be ok
for storing those objects that very rarely get updated in a working
image, methods, classes, static images etc.

Of course, we're holding our breaths for MRAM to become practical and
then lots of new possibilities open up. Go, Almaden Labs!

tim

-- 
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
Useful random insult:- Teflon brain -- nothing sticks.  -- Lilly Tomlin




More information about the Squeak-dev mailing list