On Mar 22, 2009, at 8:21 PM, Vitaliy wrote:

> You worked on a PDP-11? Wow! :-)

Well, no.  I learned to program a PDP-11 in assembler class.
I did work on PDP10s rather a lot.


> I'm kind of curious why there doesn't seem to be (from my non-expert  
> POV) as much reuse of old proven concepts as one would expect.

We've learned things since then.  One of the things we've learned (a  
large part of the whole RISC architecture fallout) was that "elegant"  
machine-language architectures aren't necessarily what makes for a  
good overall system.  For all it's elegance, I wouldn't want to be  
thrown back on a 7-register (plus the PC), 64k machine like the PDP11  
to do actual work.  And the things added to it for more memory, or  
floating point, or whatever probably lacked the elegance (I never had  
to deal with it.)

As an obvious example, microcontrollers are expected to have  
instructions for bit manipulation in minimal space.  No such thing on  
the PDP11.


> Like, if you say that PDP-11's architecture was so great, why  
> couldn't someone take it, and build a microcontroller based on it?  
> What would be the tradeoff?

In the early days of microprocessors, EVERYONE was claiming "PDP11- 
like architecture."  The 6502 claimed such, as did the 6800, and  
almost everyone else.  AFAIK, the PDP11 originated concepts like  
autoincrement and decrement addressing, and using the PC as an index  
register to utilize "immediate" values, for example.  And of course  
DEC owned the architecture and was fumbling around with it (there was  
a single-chip LSI-11 equivalent, for example.)  By the time chip  
technology had reached the stage where a PDP11 microcontroller would  
have been practical (relatively recently, I would think; you wouldn't  
want any less than 16k flash for sure...), and the licensing was less  
encumbered (if it IS less encumbered), the architecture may have been  
"elegant, but rather quaint and old fashioned.)

The PDP10 is another example.  No one is likely to be implementing a  
36bit non-byte-addressable CPU any time soon.

Although, I continue to be surprised that someone with a lot of time  
on their hands doesn't throw up an  implementation on freecores.org or  
similar.  The PDP6 (predecessor to the PDP10) was supposed to have  
been implemented in about 3000 gates.

(I guess the OTHER issue is that although these may have been "widely"  
loved in their day, the number of people actually familiar with the  
architectures is very small compared with the number of java  
programmers, in absolute terms.)

BillW

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist