On Tuesday 18 March 2003 19:24, Alan Kay wrote:
> With all due respect, not really... The Dynabook was actually aimed
> at a MOS multiprocessor architecture, which was emulated on the Alto
> (1973 - Chuck Thacker) and realized on the Notetaker (1978 - Doug
> Fairbairn). So it was quite possible to configure reasonable machines
> with reasonable balance between high level architectures and memory
> bandwidths.

Oh, I see! Either 16 virtual processors multiplexed on a single physical 
TTL processor or 16 slow MOS processors would get roughly the same 
results. I should have figured that out since one of my early projects 
used 2 Z80 chips for exactly that reason, but I guess I was 
concentrating on the microcoding aspect of Alto and lost sight of this.

The time has come, once again, for us to think of these things. Here is 
an interesting comment by Jan Gray on what the new Xilinx XC2S600E 
means to him (see Nov 28,2002 entry at http://www.fpgacpu.org/):

    At 60 processors per $45 device (in huge volumes), that works out to
    just $0.75 per processing element. Loaded up with DRAM, this implies
    a total component cost of ~$1.50/PE, and a density of about 20-40
    processors per square inch.

> The Sony PSII and PSIII are much more interesting *and* they allow
> microcoding in many important places.

I have not seen many details about them, specially the latter.

> Well, IIRC, the first ARM that Tim did so many neat things with was
> only about 25,000 transistors (and the Alto was a lot less than
> that).

I've been there too (and have a picture - 
http://www.lsi.usp.br/~jecel/merlin4.html) but am now aiming for really 
tiny (except RAM, which is a whooping 8MB since I couldn't find any 
smaller). Implementing an Alto on the FPGA development board would be 
really fun...

-- Jecel