Re: [C--] C-- development

 > Has development on C-- ceased?  Or just stalled for a few years?

The C-- project ran out of funding around 2009.  The principals have
all moved on to other projects.  Some of the technology got
transferred to the Glasgow Haskell Compiler.  =


If C-- makes a comeback, it will probably be in the area of run-time
systems rather than code generation.  This prospect is a remote
contingency.

 > I notice in there seems to be files in qc--20090627/src with makes =

 > starting with 'mips' and 'alpha' and 'ia64'.  Does this imply that these
 > architecures are actually supported?  Or are they still under developmen=
t =

 > and stalled along with the rest of the system.

The mips got supported to the extent that it was possible to run
benchmarks on it.  I think the mips support is probably pretty good,
although I think it is limited to 32-bit mips.

 > Is  IA64 the Itanium?  =


Yes.

 > Is there other source code for C-- lying around somewhere that isn't in =

 > the destributed package?

I don't think so.   I'm open to putting the entire project on github,
but migrating from CVS to git is a nuisance.

It's possible that Jo=C3=A3o Dias has some bug fixes that have not made it
into the main repository.

 > If I were to want to use it on, say, an ARM machine, would I have to do =

 > it myself?

Jo=C3=A3o definitely got code working on the ARM, but with some limitations
that I think might have to do with floating-point computation.  And we
don't support the thumb instruction set.

 > Would it be better to continue on from qc--, or better to start from =

 > scratch?  Are there lessons to be learned from the qc-- experience I'd =

 > better find out about before I even think of  doing this (and maybe thin=
k =

 > better of it and find another worthy project?)

What problem are you trying to solve?

My opinion of the C-- language and API is that we got a lot of things
right.  The next generation of the run-time interface would have been
better, and there are probably some notes kicking around about how to
do that, but it's quite usable as it.

A possibility we overlooked was that people would want to use C-- to
write a run-time system.  There needs to be a new dialect "Systems C--"
to make that idea work well.  That project never got beyond casual
conversation.

I'm not so sure about the compiler.  It would up being first and
foremost a tool for doing compiler research.  It's a very good tool
for that purpose, but there's a lot of abstraction and the compiler is
hard to learn.  Also, the internal data structures are not as good as
the refined versions that eventually went into GHC under the Hoopl
badge.  On the other hand, none of the code-generation ideas got into
GHC, so you picks your money and you takes your chances.


Norman
_______________________________________________
Cminusminus mailing list
[email protected]
https://lists.eecs.harvard.edu/mailman/listinfo/cminusminus