CPU running smalltalk bytecode

Tim Rowledge tim at sumeru.stanford.edu
Fri Feb 8 16:47:31 UTC 2002


Hannes Hirzel <hirzel at spw.unizh.ch> is claimed by the authorities to have written:

> > Also, I have heard Lex talking about a CPU able to run Smalltalk platform.
> > Is it possible to have more information about it ? Perhaps next generation of CPU (Motorola, Intel, AMD, ...) could include such features...
Several attempts have been made over the years; Sword-32, Objektiv, SOAR
etc. The main conclusion one can draw is that there is no good evidence
to make it seem worth the effort. The chief problem is essentially
economics - in the time it takes to develop a special chip the chaps
working on the general puropse one have got a couple of generations
ahead and are twice as fast as your special. I imagine that if you could
truly apply as much design and production oomph to the task as Intel
does to their x86 monsters then you might possibly beat this problem.

However, you then run the risk of locking yourself into an architecture
that you can now see to be sub-optimal. For example Sparc has (had?)
some tagged integer arithmetic instructions that were pretty much ther
to support VW. Unfortunately they turned out to not be quite right for
thepurpose and didn't get used. Waste of money, time, effort in the end.

My feeling is that a really simple RISC cpu is likely to be most useful,
long with a really fast memory system. I've pointed out many times that
there are three important metrics fora fast Smalltalk system - memory
bandwidth, memory bandwidth and memory bandwidth. Give me a 600MHz ARM
10 and 600 MHz 128bit wide MRAM and watch the bits fly :-) A simple RISC
isa is going to be easier to write and maintain a translator for and
thus it will probably get to work better.



> >Intel was finishing up the iAPX-432, a 32-bit object-oriented "mainframe
> >on a chip", and contracted with me to build a Smalltalk system ("OPL
> >432") for the 432 evaluation board product.  
> 
> It would be an idea to track the papers on this iAPX-432 and try to find
> out which were the reasons that it wasn't a success.
There is at least one paper i the GreenBook (Bits of Histoy, Words of
Advice) which details an attempt at this. One of the problems was that
the chip never really worked up to par.

tim
-- 
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
The computer is mightier than the pen, the sword, and usually, the programmer.




More information about the Squeak-dev mailing list