Python CPU

[Stephan Houben]

Eugene Leitl <eugene.leitl at lrz.uni-muenchen.de> writes:

> Have there been any attempts in creating an ASIC capable of
> interpreting Python opcodes directly? 

I don't think so.

A similar attempt has been made Very Long Ago with the Lisp Machine,
the idea being that conventional architectures are optimized for
C-like languages and that Lisp would benefit from a specialised 
architecture. 

However, conventional architectures improved so fast that soon
Lisp interpreters running on them were faster than a hardware
Lisp implementation. 

I think a similar story can be held for Python; it might be
possible to build a Python Machine, but making it as fast
as a Python imterpreter on a i486 will be *very* difficult.

Moreover, what do you do with all those C extension modules?
Anyone volunteering to extend gcc with a Python-opcode backend?

> Of course, with 
> soon-to-arrive Java chips and JPython there might be not 
> that much need for it...

I guess that Sun is relearning the lesson of the Lisp Machine
the hard way...

Greetings,

Stephan