From: Godmar Back (gback@cs.utah.edu)
Date: Tue Jan 05 1999 - 17:28:22 EST
This URL calls itself "the" Garbage Collection Bibliography.
http://www.cs.ukc.ac.uk/people/staff/rej/gcbib/gcbib.html
One of the things I saw right on the first page is Agesen's and
Detlefs's paper on how to compute layout maps for precise collection in Java:
ftp://ftp.dcs.gla.ac.uk/pub/drastic/gc/detlefs.ps
We will have to implement what they describe there at some point.
Thinking about how much work is still left on Kaffe, I think that we
must make our modularization efforts a top priority --- GC and JIT
spring to mind.
I do believe that with a well-defined & documented GC & JIT
interface, we would be able to attract the resources needed to bring
them closer to the state of the art. For example, look at TYA: these
guys were able to write a complete JIT for Sun's JVM without even
looking at Sun's code. Or ShuJIT. (BTW, is there any chance we could
support the java.lang.Compiler interface?)
Similarly, the people who've tried to implement a better gc for
kaffe (like Wrumpelnik in his thesis) had a hard time because of the
lack of a defined interface.
I think we're making progress, though, although it's a bit chaotic
and sometimes I feel either like tapping in the dark or like rediscovering
things that other people have already discovered --- but hey, that's also
part of where the fun is.
- Godmar
This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 19:57:29 EDT