SL Baur writes:
The area that we're still weak(est) at, performance wise, IMO,
is
startup time. At least it's the most noticeable.
Talk to Mike, the stat guy. The main slowdown is in finding the
packages which cost a lot of stats.
Do we still have anyone doing performance benchmarking? We used to
have
a test suite for that.
No. Martin and Hrvoje used to do that stuff with the Lisp
interpreter, but those don't seem to be as important as stuff like
lexical scope; all that bondage and lack of discipline is very costly
in CPU cycles, I think. There are also nasties (infloops, maybe?) in
the regexp code.
The other place where we're really hurting in my somewhat informed
opinion is in redisplay. The widget work added loads of additional
dirty flags resulting in a lot of X traffic, a significant fraction of
which are not #ifdef'd for widgets so they slow things down regardless
of whether widgets are in use or not.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta