If I delete all the [clip] objects in the attached patch, it becomes much faster to select and move the toggle objects. Why is that? I mean, the number of selected objects is the same either way. Do the [clip]s get redrawn every time I move the selection?
-Jonathan
On Wed, 2010-11-24 at 20:30 -0800, Jonathan Wilkes wrote:
If I delete all the [clip] objects in the attached patch, it becomes much faster to select and move the toggle objects. Why is that? I mean, the number of selected objects is the same either way. Do the [clip]s get redrawn every time I move the selection?
-Jonathan
Not sure. It really bogs down on my machine but it is still way faster than the older algorithm. To see just how many messages pass between pd and tcl/tk run:
pd -d 1 lots-of-objects.pd
Compare old version vs. new and you'll see the difference.
The ironic thing is that I also noticed some inconsistencies in expected performance. Namely, when I have one complex GOP, it runs relatively sluggish. Then I add the second one and everything moves *way* faster (GOPs BTW are not accelerated in my code either). It is almost as if at some point threshold on the socket queue elevates its priority above other things (but this is a mere speculation at best as I've no idea how kernel handles these kinds of things, particularly when one is running a rt kernel).
Best wishes,
Ico