On Feb 10, 2010, at 12:13 PM, Sean Luke wrote:
It looks like it was an Apple java bug: I just did another upgrade to an even more "latest" java release and the problem went away.
Was definitely not GC, I'd checked for that; and things ran great when using 3D libraries, go figure. Looks like it was a bug in Apple's graphics queue.
I've been working with a new 3D graphics framework in eclipse (GEF3D) and it is consistently faster for 2D graphics as well. They're actually promoting it as an alternative for graphical editors and it works really well. I guess taking advantage of all of those transistors in 3D chips pays off. Which justifies OpenCL, etc..
Did you have to do a manual upgrade? I'm using 1.6.0_17.
On Feb 10, 2010, at 2:49 PM, Miles Parker wrote:
I haven't seen these in my own non-MASON API code using SWT. (That's not intended as a snarky comment, just an extra data point. :D). But I don't actually think it's AWT because I just ran some (pretty simple) Ascape code to see and I asn't seeing this either. Are these particularly large models? When you say 2D I assume you mean that there aren't any 3@ libs involved.. I'm using latest VM under Snow Leopard running under pure 64 bit mode. Two thoughts that probably have already occurred to you..
1. This sounds like classic GC hyperactivity. Have you taken a look with YourKit and/or futzed with heap and permgen? If you tweak the gc you should at least be able to see if the policy affects the behavior. Do you have a lot of weak references or anything like that?
2. I've forgotten what the other one was.. :) Oh yeah, 3D libs but again it doesn't sound like you're using those.
On Feb 10, 2010, at 8:21 AM, Sean Luke wrote:
I've noticed that when running under the latest version of Java on OS X, simulations will run fine under 2D visualization, then suddenly drop to half their speed for a few seconds, then jump up to full speed again, and back down, and so on. I'm trying to nail down what might be the problem. I regressed to very early versions of MASON and they're all having the problem so I suspect it may be something tickled by Apple. However I can't reproduce it with simple animation tests outside of MASON. Has anyone else seen this?
Miles T. Parker
President and Chief Software Architect