[gclist] Re: "This Old Space", with apologies to Will Clinger

Nick Barnes nickb@harlequin.co.uk
Tue, 10 Dec 1996 11:47:35 +0000


> Henry Baker's example illustrated the superiority of my
> 2-generation "non-predictive" collector over conventional
> generational and non-generational collectors for the radioactive
> decay model.

1. Do you account for the cost of finding roots when collecting the
older generations? Henry Baker's illustration does not.

2. I think the lesson we should all be learning here is that
radioactive decay is often a very poor model indeed.

> Mark Tillotson mentioned the following hypothetical example:
> >         Total live heap:            30MB
> >         Spare memory:                5MB
> >         youngest generation size:  0.1MB
> >         allocation-rate:             1MB/s
> >         promotional-rate from youngest gen: 0.005MB/s
> >           (I'm positing 0.5% survival rate after 100K of allocation)
> 
> These numbers are statistically impossible for the radioactive
> decay model.

But are altogether realistic in practice.

> Let me repeat Henry's caveat:  Real world heaps have statistics
> very different from those of the radioactive decay model.  The
> radioactive decay model is primarily useful for Gedanken
> experiments that improve our theoretical understanding of
> generational garbage collection.

It seems we would do well to find some other simple model, more
realistic than radioactive decay but still simple enough to be
tractable. I have a complex model, but not a simple one.

Nick B