Avi Bryant <avi at beta4.com> wrote:
> On Sat, 25 May 2002, Bob Arning wrote:
> 
> > There may be several factors that add up to the difference you see.
> > One of them is that in 2.8, selecting a class did NOT select a method
> > category and thus showed an empty message list. In 2.9, the '-all-'
> > category gets selected by default and this causes the message list to
> > be populated. Bigger classes take longer to do this.
> 
> <sheepish>That would explain that, then</sheepish>.
> Ok, so that benchmark was useless.  However, the slowdown is still there
> (honest!); I'll have to find a cleaner way of demonstrating it.



The problem is real regardless of the cause.  Maybe we shouldn't select
--all-- by default if it's going to be slowish.

Incidedentally, my "large lists" patch would probably help here.  I
added it to help with very large lists in Celeste and (if I ever got
around to it) in IRC.  It should help with the display of any large
lists, however.

Speaking more generally, speeding up common operations like these seems
to be much more promising than doing things like speeding up BitBlt or
DamageRecorder.  For another example, I once axed a delay in the opening
of the main morphic menu, by finding a way for the current project to be
located more quickly.  (And then Dan I. found an even faster way). 
(Incidentally, that menu is pretty slow again as of changeset 4843!)


Lex