Are you able to build Pivot from source? If so, can you try commenting out the 
calls to paintVolatileBuffered() and paintBuffered() in 
ApplicationContext.DisplayHost#paint()? That should tell us if the double 
buffering is the problem.

On Nov 16, 2010, at 4:12 PM, Mike Smorul wrote:

> I spoke a little too soon about using -Dsun.java2d.pmoffscreen=false,
> it was a webstart test, so I needed to specify
> -J-Dsun.java2d.pmoffscreen=false. Interactive performance improved,
> however bandwidth requirements were just as severe.
> 
> I'm not sure if java has a clean to detect if its running remotely way
> other than checking the DISPLAY environment variable.
> 
> -Mike
> 
> On Tue, Nov 16, 2010 at 3:37 PM, Greg Brown <[email protected]> wrote:
>> It's probably because all drawing in Pivot is double-buffered, so bitmaps 
>> are being sent over the wire rather than graphics primitives.
>> 
>> We could potentially provide a way to turn this off - or maybe there is a 
>> way to detect a remote X display and automatically turn it off?
>> 
>> On Nov 16, 2010, at 3:19 PM, Mike Smorul wrote:
>> 
>>> I've been trying to run a pivot app forwarded over SSH/X11 and the
>>> performance makes it almost unusable. Screen redraws take several
>>> seconds and mouse events are delayed enough that double-clicks almost
>>> never register. A similar swing app performing similar actions (file
>>> browsing) responds as if it's running locally.
>>> 
>>> The two machines are connected via a gig link w/ .3ms latency. One
>>> major difference I've noticed betwee the swing and pivot app is the
>>> amount of network traffic. During redraws, the pivot app will push
>>> almost 2MB/s while the swing app hardly ever pushes over 100KB/s. I've
>>> tried the usual java2d trick of '-Dsun.java2d.pmoffscreen=false', but
>>> this had little effect.
>>> 
>>> Any ideas?
>>> 
>>> Thanks,
>>> -Mike
>> 
>> 

Reply via email to