>> Cocoa SWT just can't seem to get it all together on the performance front.
> Out of interest; do you know what the issue is there? Are they using GC in the Objective-C side or reference counting?
it's not clear .. there are several bugs that explore wtf is going on:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=285175
https://bugs.eclipse.org/bugs/show_bug.cgi?id=282229
are a couple of the nastiest ones, and i SUSPECT fixing these will fix a lot of the underlying problem.
the entire control stack is wrapped in java, though, which means going in and out of native code a LOT. it's possible that just the model for this is just going to be very hard to optimize. there's a reason that carbon is the preferred implementation for systems like SWT ... it's ust way easier to do the plumbing in the carbon model than the cocoa model. look at the stack depth of the jprofiler profiling run i did on one of those bugs. it's insane. it's like 150 frames deep to repaint during scroll.
there does appear to also be a bunch of duplicate work going on, but it doesn't seem to account for enough time to explain the performance.
ms
This archive was generated by hypermail 2.0.0 : Wed Apr 07 2010 - 16:28:10 EDT