[gclist] region inferencing

Lars Thomas Hansen lth@ccs.neu.edu
Thu, 06 Nov 1997 15:37:30 -0500

>I was actually wondering if you could mix pseudo-region analysis with 
>garbage collection.

In a talk he gave at BBN last week, Jeff Siskind, author of the STALIN
Scheme system (which is not interactive and can compile Scheme programs
by making a closed-world assumption), indicated that he does indeed
combine a form of region analysis with garbage collection: objects are
allocated in different regions depending on a conservative estimate of
their lifetimes, and the most general of these regions is also
garbage-collected.  In some instances, there is a "region" for each
procedure.  No other details were given, but it was my understanding
that these features are part of the upcoming 0.8 release of STALIN.