[gclist] Name that hypothesis
Henry G. Baker
Wed, 4 Dec 1996 07:49:35 -0800 (PST)
> > > I use "generational hypothesis" for this:
> > >
> > > "Most objects die young".
> > As I have tried to explain many times, this statement is ambiguous/ill-defined.
> > See below for clarification:
> > ftp://ftp.netcom.com/pub/hb/hbaker/YoungGen.html (also .ps.Z)
> Having read this, I suggest that the generational hypothesis could be
> better stated like this:
> "A young object is more likely to die soon than an old one".
> I have previously measured this (in SML/NJ and Harlequin's MLWorks)
> and it is certainly true for all the large programs I have studied.
> You correctly point out that "most objects die young" is an inadequate
> statement of this. I think this is the result of sloppy expression,
> rather than fundamental misunderstanding.
> Nick Barnes
I still don't think this statement is precise enough. What I was
trying to point out in my paper was that objects which are more
numerous are more likely to die (in absolute terms) than objects which
are less numerous. If the 'null hypothesis' is that the likelihood of
dying is completely independent of how long you have lived already,
then the 'generational hypothesis' is one that says that young objects
die even more often than would be expected by this null hypothesis.
Now second order effects -- e.g., cache locality -- might allow you
to take advantage of even the 'null hypothesis', but to win better than
this you will need the 'generational hypothesis'.