[MLton] Welcome to John Dias

Matthew Fluet fluet@cs.cornell.edu
Wed, 12 Jan 2005 09:27:25 -0500 (EST)


> C--'s main compiler is O'caml based, used to have a SML compiler that
> used MLRisc, and appears to have some kind of long term roadmap of
> becoming the target backend for the GHC Haskell comp.

I believe that is correct.  The C-- lead (Norman Ramsey) recently had a
technology exchange with the GHC lead (Simon Peyton-Jones), which appears
to have been positive, though becoming the GHC backend is still a longer
term goal.

> If MLton gets a C-- backend the emitted code can be compiled by the
> O'Caml C-- compiler. Incestuous.

I beleive that it should be possible, although C-- is missing the ability
to create, recover from overflow, and move stacks, which makes it almost a
non-starter for MLton.  There is also the question of whether or not the
complication of the GC (which would likely lead to longer pause times) is
outweighted by the potential for additional native backends.

> I'm curious as to what the core team thinks of Moby the language.

I've not seen a good "A-ha" example of a Moby program that convinces me
that I can do something with significantly more ease than in another
language.  To be honest, I think it mixes too many different features,
which simply makes it harder to choose how to represent one's
abstractions.

> Moby's compiler is in SML, uses MLYacc and MLlex and the MLRisc backend,
> is also SML and I think someone last year successfully used MLton to
> compile the Moby compiler and MLRisc emitter. Matthew? However I could
> not find it in the archives.  Mobyton :)

Yes, I had ported MLRISC and Moby:
http://mlton.org/pipermail/mlton/2004-February/025040.html