"Ask ATI" dev thread

[TobiasTheViking]

You are just saying that to make me happy...


But please, do go on

[zim2dive]

I did a quick search of the thread and didn't seem to find this..

ATI (IMO) has picked some really poor defaults for overscan.. my 1920x1080 LCD, connected over HDMI, was defaulting to a display area that was ~1776x1000, even tho I picked 1920x1080 in the resolution settings in CCC.

The Windows version of CCC has overscan controls, none in Linux.

a) when will linux CCC be brought to parity with the Windows version in terms on controls it offers? (and isn't overscan a pretty basic function to include, esp given the choice of default?)
b) why would you pick such a large overscan as the default for a digital video output?!?!

It makes quite an initial (negative) impression, especially b/c the commands (via ati-config) are not the simplest (the "tv" overscan commands look more applicable to someone new to ATI, and they have no effect)

thanks,
Mike

[bridgman]

I suspect the underscan amount was picked to match the overscan amount that some TVs were using by default, to ensure that the user could always see the menu bar. We definitely need an easy way to turn the underscan off though... default underscan is a real convenience for new users but only if you can turn it off.

[Nexus6]

The ability to pop a new tab in IE by clicking on the stub tab at the right of the active windows is really nice and I use it a lot, which is a point in favor of IE/Vista over Firefox/Linux.

Double left-click in an empty part of the tab bar on Firefox 3.x. I'm pretty sure that's configurable too.

[rainbyte]

Hi Bridgman! I have a little question that appeared on the IRC channel while I was chating with cxo...

Would it be posible to implement some kind of closed source library (like the intel_hal.so) with the opensource driver? AMD/Ati could put all the effort in this library instead off fglrx and also this would speed up the open source driver development.

Personally, I think that and AMD/Ati would benefit because the driver will be compatible with all distros, very stable and you cuold implement special top secret features (like 2D and 3D Optimizations for workstation users) into the binary library (which would be optional to the normal user, except if they want to use UVD). This would be like the libva structure, an opensource api with a binary library.

I have been thinking about this from the first time I read this article on wikipedia: http://en.wikipedia.org/wiki/Intel_GMA#intel_hal.so

Thanks for all your attention and support to the opensource community...

EDIT: This topic is also being deliberated on the Open-Source AMD/ATI Linux forums in this thread

[bridgman]

I'll answer over in that thread; this does come up from time to time.

[zx2c4]

Resizing windows with compositing enabled (KDE4 desktop effects) is, as you know, horribly slow. Any prospect of fixing this by 9.2? Any known workarounds?

[tuxdriver]

Resizing windows with compositing enabled (KDE4 desktop effects) is horribly slow.

It's slow even on the most lightweight non-composited desktop. Looks like moving and resizing windows is accelerated poorly or not accelerated at all.

For example, try to move a large window on a high-resolution desktop (2048x1536 @75Hz / 22" CRT) or (2560x1600 @60Hz /30" LCD) and you'll see how slow it is.

There's no such problem on Windows XP,Vista or Windows 7.

[RealNC]

I can't confirm. With no compositing, it's somewhat fast (but still slower than Vista or XP). Only with compositing it's slow as hell (window contents are updated with 1FPS or such.)

[tuxdriver]

I can't confirm. With no compositing, it's somewhat fast (but still slower than Vista or XP). Only with compositing it's slow as hell (window contents are updated with 1FPS or such.)

Have you tried it with a single core CPU? Dual and Quad cores can cope with the high CPU load.

On Windows it uses only 0-2% CPU.On Linux it uses about 60-100%