[MLton] callcc and space use

[Florian Weimer]

* Vesa Karvonen:

> Just curious, since you specifically mention using callcc to suspend
> and resume computation (rather than to fork computation), I wonder
> whether there is a reason to use MLton's callcc
> (http://mlton.org/MLtonCont) rather than MLton's threads
> (http://mlton.org/MLtonThread), which are expressive enough for
> suspending and resuming computation.  I ask about this, because,
> AFAIK, MLton's callcc and throw are implemented on top of MLton's
> threads by copying the stack and performing thread switches.  IOW,
> using just MLton's threads (i.e. switch) for suspend and resume would
> likely be considerably more efficient.

I think it's the other way round.  But new threads are copied from a
specific thread with an empty stack, which is faster than copying the
current thread's stack.  However, the function to call is passed to
the nascent thread in a global variable (with suitable locking), and
this results in unnecessary contention in a parallel environment.

It seems that there is some room for improvement here.

-- 
Florian Weimer                <fweimer at bfk.de>
BFK edv-consulting GmbH       http://www.bfk.de/
Kriegsstraße 100              tel: +49-721-96201-1
D-76133 Karlsruhe             fax: +49-721-96201-99