[gclist] region inferencing

Taura Kenjiro tau@is.s.u-tokyo.ac.jp
Fri, 7 Nov 1997 12:11:30 +0900 (JST)

 > To spur some discussion on the gclist I ask: Will region inferencing
 > replace garbage collectors, supplement garbage collectors, or be
 > consigned to the dustbin?

The ML Kit version 2
(http://www.diku.dk/research-groups/topps/activities/kit2) implements
region inference. It comes with an introductry paper that informally
explains what region inference can do.

Although I haven't tried it yet, the introductry paper says that the
current region inference system sometimes needs certain coding styles
to reclaim space promptly enough. Often, objects with different
lifetimes go into the same region and anything in that region cannot
be reclaimed until all objects in the same region die. For
memory-demanding programs, programmers may have to have some knowledge
about what the inference system does at compile time and work around
it.  The behavior of garbage collected system (in particular, ones
that meet "safe-for-the-space-complexity" rule), on the other hand, is
more easily understood and predicted by the programmer.

Of course, I'm merely talking about the particular region inference
system currently implemented, based on my current intuition and
understanding. There will be future improvements that make region
inference more transparent to the programmer. My feeling: it is
already worth considering as a supplement of GC and as a research
topic :-).

	Kenjiro Taura
	Yonezawa Laboratory
	Department of Information Science
	Faculty of Science
	University of Tokyo
	7-3-1 Hongo, Bunkyo-ku Tokyo 113 Japan  	
	tau@is.s.u-tokyo.ac.jp (http://web.yl.is.s.u-tokyo.ac.jp/)
	tel: +81-3-3812-2111 ext. 4108 or +81-3-5800-6913
	fax: +81-3-5689-4365