Apple

On Nov 4, 2007, at 5:42 PM, Mike wrote:

This also fixed the sluggishness in GIMP. It still crashes though. But, still whatever is causing the sluggishness in rdesktop is most likely the same problem. Now I have to go onto google and try and figure out what the heck beam synchronization is. Anyway, just thought you all might think it was interesting.

That's really interesting, as it points to X11 being the culprit rather than its clients.

It doesn't necessarily mean that it will be fixable without any client change, but it

becomes more likely.

Beam synchronization causes app that draw "too much" to be paused until the

next screen update. See <http://developer.apple.com/qa/qa2001/qa1236.html#TNTAG9> for an introduction,

and <http://developer.apple.com/technotes/tn2005/tn2133.html> for more details.

In particular recommendations for Quartz/CoreGraphics clients are at

<http://developer.apple.com/technotes/tn2005/tn2133.html#TNTAG3>.

-mike

On Nov 4, 2007, at 6:21 PM, Jesse wrote:
[...] "Disable Beam Synchronization" has returned rdesktop performance to what it was under Tiger. [...]

Thanks for posting the results of the experiment.

Are there going to be any ill effects anywhere else in Leopard by having the "disable" option turned on?

Some tearing will occur for apps that draw while the screen is being updated.

Badly behaved applications might end up drawing too much, thereby increasing

CPU/GPU load and power consumption.

Also, does this persist across uses of Quartz Debug? Actually, it looks like I have to have Quartz Debug open in order for the "disable" option to be active. I can live with this for some time if required.

I believe that quitting Quartz Debug reverts to the default behavior.

The "last resort" measure listed at <http://developer.apple.com/technotes/tn2005/tn2133.html#TNTAG9>

might help. I don't know which app should get its plist modified.

You could try adding that to /usr/X11/X11.app/Contents/Info.plist :

       <key>CGDisableCoalescedUpdates</key>
       <true/>

Would you still like the Shark profiles?

Sure. Ideally I'd prefer to see them attached to a Radar.

If there is an easy way to reproduce this that doesn't involve a complex install, I could do a bit

of profiling to give a hand to Ben in tracking this issue. I am no Core Graphics expert but I wouldn't

mind learning a bit more. Filing a bug would help to track this issue. I have access to Radar, so

<http://bugreporter.apple.com/> is great for me. It may also help Ben when it comes to argue

that fixes should be released in an Apple update. If you don't want to go through that route,

tracking this at X.org is also doable. Please post the Radar # or bug ID once/if you file.

Thanks - Eric

References:
	>speed of window 'refreshes' within X11 windows (From: Jesse <email@hidden>)
	>Re: speed of window 'refreshes' within X11 windows (From: "Nathaniel Gray" <email@hidden>)
	>Re: speed of window 'refreshes' within X11 windows (From: "Chris Linstid" <email@hidden>)
	>Re: speed of window 'refreshes' within X11 windows (From: Juan Orlandini <email@hidden>)
	>Re: speed of window 'refreshes' within X11 windows (From: Jesse <email@hidden>)
	>Re: speed of window 'refreshes' within X11 windows (From: Eric Gouriou <email@hidden>)
	>Re: speed of window 'refreshes' within X11 windows (From: Jesse <email@hidden>)
	>Re: speed of window 'refreshes' within X11 windows (From: Mike <email@hidden>)