On 13 Sep 2006, at 18:42, chromatic wrote:
Avi Bryant and Piers Cawley have some thoughts about improving
method dispatch
speed for the common case, based on ideas in Sun's Strongtalk and
the Self
language:
http://www.bofh.org.uk/articles/2006/09/13/what-we-have-here-is-an-
opportunity-to-accelerate
http://smallthought.com/avi/?p=16
What they describe is effectively branch prediction for method
despatch. Instead of looking up the right place to go and then going
you go to the most likely place and then check that you did the right
thing.
Is this similar to (or even the same as) PIC?
It's not quite the same is it? The method Avi and Piers describe
involves a normal call to a method the first action of which is to
find out whether it's really the correct method.
--
Andy Armstrong, hexten.net
