[MLton] approaches to parallelism in ML

[Kip Macy]

>
> > How do mlton developers feel about supporting same address space
> > parallelism in mlton?
>
> I think it is important to get ASAP.  Although skaller is right that
> we have to get the 64-bit port done first.  But that's coming.

Do you have any sort of realistic ETA? How much work will it be to
extend the NCG?

>
> I think the way to go is to first build a simple, non-concurrent,
> stop-all-threads collector as a variant of MLton's current collector.
> That should allow us to get up and running fairly quickly and to
> handle certain kinds of apps, possibly even a good fraction of them.
> This kind of collector was explored long ago for SML (both Poly/ML and
> SML/NJ) running on SMP machines, see:
>
>   Adding Threads to Standard ML
>     http://mlton.org/References#CooperMorrisett90
>
>   A Distributed Concurrent Implementation of Standard ML
>     http://mlton.org/References#Matthews91
>
>  Procs and locks: a portable multiprocessing platform for Standard ML
>  of New Jersey.
>     http://mlton.org/References#MorrisettTolmach93
>

Thanks. I'll take a look at these.