[MLton] More on parallelism

[skaller]

On Tue, 2006-12-05 at 11:57 -0500, Suresh Jagannathan wrote:
> We've been focussed on two issues.  The first deals with support
> for CML on multi-cores and follows the outline Matthew suggested.
> Essentially, we have a 1-1 mapping between OS level threads and
> MLton schedulers, on top of which CML threads are scheduled.
> Currently, we simply partition the heap uniformly among all
> OS threads for purposes of allocation; all threads stop when any
> one triggers a GC.
> 
> However, while this strategy makes sense for tightly-coupled
> clusters,

It does? Felix does this too. I doubt it makes sense to 
stop 64 processors to do a collection. It seems particularly
bad if some processors might take a while to enter the GC,
the delay then depends on the one that takes the longest.

Basically, one uses multiple processors only to improve
real time performance .. so stopping the world doesn't
seem even marginally reasonable to me.

-- 
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net