In article <······················@slip-80-6.ots.utexas.edu> ·····@clyde.as.utexas.edu (James McCartney) writes:
In article <··········@cerberus.ibmoto.com>, ···@apple.com (Tim Olson) wrote:
> Garbage collectors usually divide memory into two sets: unreachable
> (provably unused) and reachable (not provably unused). Is anyone aware
> of any work that has been done in additionally performing some sort of
> aging on the reachable set? What I'm thinking of here is GC support
> for knowing what objects are recently used vs. not recently used, so
> that space-time tradeoffs can be performed automatically (e.g. JIT
> compilation of frequently-used functions, discarding generated code for
> infrequently used functions, or compression of infrequently-used
> objects).
It would seem that to keep track of that info you'd have to have
some extra code executed at each access which would be expensive.
Wouldn't you have to do this (or something like it) for any
just-in-time compilation mechanism?
Ajit