View Full Version : Opensource ati driver - Ubuntu 8.10/9,04 difference
tepegoz
05-05-2009, 07:18 PM
Hi folks,
I believe this is the correct place to ask my question.
I am using an IC design program, called cadence, on both Ubuntu 8.10 and 9.04. My graphic card is Ati Radeon HD 3650. I am using open source ati driver. If someone asks why not binary drivers, that is a long and sad story, I don't want to go in that. Shortly, ati binary drivers are problematic for this hardware. But open source drivers are just fine except 3d effects and hardware accelerated opengl.
So what I will ask is this: There is a certain difference between the open source drivers of Ubuntu 9.04 and 8.10. For an illustration I captured 2 screens from both 8.10 and 9.04. See them in this link:
http://www.mems.eee.metu.edu.tr/~tepegoz/ati/
screen_intrepid.png shows the case for ubuntu 8.10 and screen_jaunty.png shows the case for ubuntu 9.04. screen_intrepid is what it should be. as seen, on screen_jaunty, there are many misdrawen areas on especially vertical lines.
So the ati opensource driver on ubuntu jaunty is problematic for my case. Do you have any idea how to correct the driver on jaunty? Or what is the main reason for having such a difference?
Regards and thanks everyone in advance
bridgman
05-05-2009, 10:42 PM
The main difference between Intrepid and Jaunty for a 6xx card is that the drivers in Intrepid used shadowfb acceleration (software rendering to a frame buffer in CPU memory) while the drivers in Jaunty used hardware acceleration for 2D and Xv.
Check in your xorg.conf for Option AccelMethod EXA, and if it`s there comment it out and restart X (eg log out and back in).
I`m not sure if that will be enough to force the driver back to shadowfb, but if that doesnt work try Option DRI Off
nanonyme
05-06-2009, 03:38 AM
The main difference between Intrepid and Jaunty for a 6xx card is that the drivers in Intrepid used shadowfb acceleration (software rendering to a frame buffer in CPU memory) while the drivers in Jaunty used hardware acceleration for 2D and Xv.
Check in your xorg.conf for Option AccelMethod EXA, and if it`s there comment it out and restart X (eg log out and back in).
I`m not sure if that will be enough to force the driver back to shadowfb, but if that doesnt work try Option DRI OffIf it's a recent version of the driver, it will enable DRI if the device node exists and then enable EXA if DRI was set on. I'd strongly suspect blacklisting the radeon driver (preventing it from loading) would at the very least give shadowfb but it might be urgent to investigate whether that which looks to me like clearly a bug still happens with up-to-date drivers. (read: git master of xf86-video-ati)
Edit: Might still be possible to prevent it from loading with variables, that blacklisting just should equal a scenario with too old a kernel for DRM support.
Ren Höek
05-06-2009, 06:14 AM
[..]
Check in your xorg.conf for Option AccelMethod EXA, and if it`s there comment it out and restart X (eg log out and back in).
[..]
AccelMethod EXA is the default option in jaunty, so you have to set the option
Option "AccelMethod" "XAA"
to go back to XAA as in intrepid (desktop effects are active in default as well).
Sorry if I misunderstood the comment, but looked like this wasn't clear.
monraaf
05-06-2009, 08:08 AM
Looks like a bug to me, similar to the one with audacity displaying corrupted curves.
http://bugs.freedesktop.org/show_bug.cgi?id=21561
You should probably also file a bug report.
tepegoz
05-06-2009, 08:33 AM
Thanks for the comments folks,
I will change AccelMethod as soon as possible and give you feedback.
Regards
bridgman
05-06-2009, 09:51 AM
One minor point - XAA acceleration was never implemented for 6xx/7xx. Since there is no legacy 2d hardware, we went straight to EXA since the EXA operations are a better fit for the 3D engine. As a result, the valid choices are shadowfb or EXA.
Since EXA acceleration had not been implemented in the Intrepid timeframe I'm assuming that Intrepid used shadowfb.
I don't think the radeon driver accepts "shadowfb" for accel method so disabling DRI somehow is probably the best bet.
agd5f
05-06-2009, 12:12 PM
Option "DRI" "False"
is what you want. that will enable shadowfb for r6xx/r7xx
tepegoz
05-06-2009, 06:23 PM
Ok folks, here is what I got on Jaunty:
1. I tried Option "DRI" "false" and that weird misdrawing disappeared. However, window drawing speed is very poor, even the window manager is not usable.
2. I tried Option "AccelMethod" "XAA" and this even did not work. I could not get X running.
So, it is still weird for me that, on intrepid I got fast and correct drawings. When I disable DRI on jaunty, yes, the drawings become correct, but the system speed becomes terrible.
So, does this mean that there is something wrong with the open source driver on jaunty?
I will stick with intrepid until I can solve this problem.
Thanks and regards to all
bridgman
05-06-2009, 06:51 PM
Hmm. Sounds like maybe the driver is not falling back to shadowfb like it should.
vBulletin® v3.8.4, Copyright ©2000-2009, Jelsoft Enterprises Ltd.