[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [moca] Locality and Name matching in Pi-Calculus.

Dave Turner's PhD thesis

   http://www.cis.upenn.edu/~bcpierce/papers/pict/turner-thesis.ps.gz

showed that you can get a surprisingly good compiler for the functional part
of an ML-like functional language by going via pi-calculus.  You'd think
that the translation to pi would throw away a lot of information that is
needed in the back end, but it turns out that a little bit of cleverness in
the pi-level type system is enough.


well, given that we now
know how to translate a large number of high-level
languages into pi fully-abstractly (and there is good prospect that more
can be done along these lines), we don't throw away any (semantic) information
if we set up the translation in the right way. [as an aside: my experience
with encodings even suggests that we don't throw away complexity information,
but i have no way of formalising this intuition, alas.]

that leads me to think compilation via pi should essentially produce
the same code as a direct translation. whether this is the case or not,
something interesting appears to be suggested:

1 if it didn't, i guess the culprit would be the (pi --> assembler)
  part. That means we should look harder at compiling pi.

2 if it did, that would seem to indicate that the existing compilation
  process for pi is essentially optimal, or at least as good as it can
  be at this point, given the state of the art in conventional compiler
  technology.

my hunch is that (2) holds, but there's a caveat: almost all investigative
effort has been spent on translations of sequential high-level languages.
there are various reasons for this restriction, most of them non-technical,
but i wonder if the picture would change if we'd also take into account
the translation of concurrent HL languages, if indeed there is such a thing.

martin


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The "models for mobility" mailing list mailto:moca@xxxxxxxxxxxxxxx
http://www-sop.inria.fr/mimosa/personnel/Davide.Sangiorgi/moca.html