Is there a modern GC for D?
Andrei Alexandrescu
SeeWebsiteForEmail at erdani.org
Tue Feb 9 22:06:47 PST 2010
Robert Jacques wrote:
> Based on a thread on the DMD concurrency mailing list I've begun to get
> a sinking regarding the future of the garbage collector in D: most of
> the work in GC algorithms has gone into functional and (mostly) pure-OO
> languages, leaving a multi-paradigm systems programming language like D
> out in the cold. So far, I know mark-sweep and mark-region algorithms in
> either serial, parallel or thread-local forms should work. But based on
> Java we'd really like incremental, generational and/or concurrent
> options. So I'd like to ask people to help brainstorm some ideas.
>
> Some things I've run into:
> -structs/pointers/slices/etc make finding memory meta information, like
> mark-bits or ref-counts, non-trivial.
> -C function calls and assembler blocks make code instrumentation
> questionable.
> -Getting concurrent GC code correct is very hard. Boehm's algorithm, for
> instance, looks extremely racy.
>
> Thanks!
s/sinking/thinking/
http://www.youtube.com/watch?v=gmOTpIVxji8
:o)
Andrei
More information about the Digitalmars-d
mailing list