On 08/25/2009 09:16 PM, Stefan Schmiedl wrote:
On Sun, 23 Aug 2009 15:08:24 +0200
Paolo Bonzini<[email protected]> wrote:
Thanks for improving this central part of gst,
It's not improving. It's making it do what was meant to do.
Oh ... well, if that's so, then it was about time :-)
Ahah :-)
As an added bonus, I just reran my Iliad benchmarks (with the most recent
gst and iliad) and the timings are between 5% and 10% percent better.
Memory usage seems a bit improved, too, but I'm not really sure there.
Good, I couldn't have given numbers, but that would be expected since
now object allocation is faster -- after the incremental GC finished,
it's just a
while (!IS_OOP_VALID (oop))
oop++;
(Interestingly, one of my first changes to GST ever was to change object
allocation to use a linked list. This went away with the new GC, but
now that the focus on network application is bigger, the incremental GC
behavior becomes more interesting and using a linked list would make
sense again. I don't really feel like introducing even more
instability, but I have to keep this in mind for the future).
Paolo
_______________________________________________
help-smalltalk mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-smalltalk
