FGLRX Catalyst and Resizing with Desktop Effects

[RealNC]

Don't forget the heavily annoying slowness of maximizing/un-maximizing windows as well restoring them from the taskbar after they were minimized. That is really annoying as hell.

Also the plethora of other bugs of course (hang while switching users or 3D stopping to work after logging out of KDE4 and then back in...)

[storma]

Using stretch I have no lag. Minimising and restore, still no lag.

A second or so lag on maximising and restore.

[rampage7]

I don't think that'd solve the complaint of there being a lag between dropping the handle and the window being properly resized.

Exactly - I've mentioned in few places this problem about six months ago, when I've bought Radeon 4850. From these times, there was six drivers releases from Ati, and any of these drivers didn't resolve this problem.
I've even reported a bug to AMD - without any results - they're treating users as an idiots... Look at this bug report and stupid answers from AMD/Ati employees: click

In my opinion, the only hope for r600/700 owners is fast development of open sourced 2D/3D driver. Fglrx driver quality is miserable - much more than Nvidia closed source driver.

[bridgman]

Just curious, what were the "stupid answers" from the support desk ? All they said was "you're sending this to the wrong place" and offered a couple of alternatives including the recommended bug tracker (the Bugzilla database at http://ati.cchtml.com. Our developers do look at *that* bug tracker.

When you persisted they asked you to read their previous response again (in the "solution" field), ie they were saying "DUDE, YOU'RE BUGGING THE WRONG PEOPLE" as nicely as possible

I'll see if I can get Phoronix added to the list of recommended support forums; I think there may be more Linux activity here than at the sites they mentioned.

[rampage7]

Just curious, what were the "stupid answers" from the support desk ? All they said was "you're sending this to the wrong place" and offered a couple of alternatives including the recommended bug tracker (the Bugzilla database at http://ati.cchtml.com. Our developers do look at *that* bug tracker.

When you persisted they asked you to read their previous response again (in the "solution" field), ie they were saying "DUDE, YOU'RE BUGGING THE WRONG PEOPLE" as nicely as possible

Exactly - it was so much annoying... I hate closed source drivers.

[Maarte]

Perhaps try this?

http://aseigo.blogspot.com/2009/01/todays-tip-kdeskipargbvisuals.html

[elbioleg]

Another one with the same problem. I'm using Debian testing and I have an ATI EAH3850. All drivers I've tested are buggy (lag when I try to restore, minimize, maximize and scale windows). I have tried all the possible combinations in xorg (there's a thread in compizfusion forum with "hacks" to improve the driver perfomance), but they won't fix the problem.

But when I activate compiz under Ubuntu 8.04 it works perfectly: no lag, no scaling issues. What is the difference between Ubuntu and Debian drivers? If we had the chance to install Ubuntu driver in Debian/Arch/... or guess where is that difference we may solve the problem.

[dkasak]

Exactly - I've mentioned in few places this problem about six months ago, when I've bought Radeon 4850. From these times, there was six drivers releases from Ati, and any of these drivers didn't resolve this problem.

Resolving 'this problem' isn't as easy as you might think. When you trigger a resize of an application window that's using dynamically sizing widget toolkit ( such as modern Linux desktop apps do ), the application re-renders the whole window. The reason for this is that when you create a window, you place your widgets inside containers and then tell the toolkit which widgets should resize to take up available space, etc. This means that you can design an app that works well at 800x600, but when someone opens the window fullscreen at 1280x1024, the widgets will resize to take available space, and the app will look good ( as opposed to the behaviour of, say, lots of Windows apps ... think VB or Access apps ... that will just display their content in the top-left corner of the screen and everything else will be blank ).

So each pixel that you increase the window size by in the X or Y direction means the application will recalculate where everything should go, and then render it.

So if you've got poor XRENDER acceleration, you're really going to notice when you're resizing windows, because you're firing off LOTS of resize events and continually repainting the application window.

[RealNC]

Yes, and? We're talking 3GHz multi-core machines here. Not some Pentium 2 300Mhz machine. The drivers have to be written in an incredibly crappy manner to produce this slowness in today's monster machines.

There's no real excuse.

[dkasak]

Yes, and? We're talking 3GHz multi-core machines here. Not some Pentium 2 300Mhz machine. The drivers have to be written in an incredibly crappy manner to produce this slowness in today's monster machines.

There's no real excuse.

I wouldn't be surprised if fglrx containted NO acceleration for later ( r600+ ) chipsets for 2D operations. Keep in mind they dropped the 2D engine completely for the r600. If this is the case, I suppose the excuse is 'there is currently no acceleration for that'.

Now I'm not defending ATI's drivers. I'm just explaining they 'why' bit. I was one of the original ones to vote for XRENDER acceleration at: http://ati.cchtml.com/show_bug.cgi?id=7 ... bug #7. Have YOU voted?