Re: Promptness of collection

Richard Brooksby wrote:
> For what it's worth I did an experiment on MLWorks (N-generational
> bucket brigade nearly purely functional GC) where I made the compiler
> null out registers and stack slots at the end of their live ranges.  It
had
> no measurable effect on heap size or GC performance.  But that was
> ML with a very high infant mortality, and mostly running a compiler,
> and quite a long time ago.

Was the other side of that comparison retaining all references to the end of
scope or was it reusing registers and stack slots for other references?

Cheers,
Jon.