What does RR do when it encounters an error?
raney at metacard.com
Thu Apr 17 13:41:01 EDT 2003
On Wed, 16 Apr 2003 Victor Eijkhout <eijkhout at cs.utk.edu>:
> Mac OS X 10.2.5, RR 2b5.
> My application runs into an error, I think "I'll do something else
> for a while". Then I notice that RR is eating half of my cpu
> capacity. Why on earth? It's supposed to be stopped for an error.
If the dialog has a "throbbing" default button, this is to be expected
and just the result of the choice Apple has made for all of us: eye
candy always takes priority over performance. I've even read on the
Carbon mailing list that for most small applications the doc resizing
and bouncing icon during application startup actually take up more of
the CPU time than the actual application startup code! Another
example: due to window buffering, antialiasing, and the complex
"gumdrop" standard buttons (not even throbbing) it takes an order of
magnitude (that's a factor of 10, folks) longer to flip through the
MetaTalk Reference stack on OS X than it does on Linux or Win32
systems of equivalent performance.
The throbbing default button performance problem is worse in MC/RR
than in some other apps because Apple still hasn't finished the
Appearance Manager API for Carbon and our workaround is a bit more
baroque than what other apps have to do (i.e., they'll only consume
25% of the CPU time whereas MC/RR might consume closer to 50%). We'll
be able to get closer to the lower number once Apple gets around to
finishing the Carbon API, which they've promised to do, but without
specifying which OS version will have it.
Note that the performance problems are much less objectionable on new,
fast hardware, and especially on dual-cpu systems. Given their UI
priorities, one might even suspect that what Apple is really trying to
do is retroactively implement some kind of planned-obsolescence
design. If you don't want to play that game, I recommend setting the
lookAndFeel to Motif or Windows 95 which will restore at least some of
the lean-and-mean performance you enjoy on those other platforms.
> Victor Eijkhout <eijkhout at cs.utk.edu>
> tel: 865 974 9308 (W), 865 673 6998 (H), 865 974 8296 (F)
Scott Raney raney at metacard.com http://www.metacard.com
MetaCard: You know, there's an easier way to do that...
More information about the use-livecode