On Thursday 19 November 2009 07:44:42 am Matthias Hopf wrote:
On Nov 19, 09 07:05:13 -0600, David C. Rankin wrote:
I've been busy lately and haven't had time to focus on the radeonhd driver issues on my Toshiba 205d laptop with RS690M (x1200). Until now, I have been
[...]
In opensuse 11.2, the screen shakes and stumbles bad enough to give you sea sickness. It looks like what is occurring is the driver incorrectly draws or corrupts a few horizontal lines (at radom vertical heights in the screen) and then the remaining part of the display gets adjusted up/down to fill the void for the remainder of that refresh (or the next 3-4 refreshes). Typing
If I understand you correctly, these issues are intermittent, meaning they happen and vanish again without you doing anything, right?
That's correct. Here a 2 more artifact screenshots from konqueror trying to copy a folder from one place to another: (~25k each) http://www.3111skyline.com/download/openSUSE_bugs/112/ss/radeonArtifacts2.jp... http://www.3111skyline.com/download/openSUSE_bugs/112/ss/radeonArtifacts3.jp...
If that is the case, it's a line buffer issue. We had more of them in the past, and there has never been a really good solution. The radeon driver has some logic to reprogram the watermark registers for that, but AFAIK it never helped completely with these issues. Interestingly enough, the R6xx/R7xx series doesn't seem to have this issue. Maybe because the memory bandwidth is just high enough...
Well that's the funny part. I have the Arch Linux install for this same box running the ATI driver and the display is fine. Also, this is an RS690M so whatever difference the 'M' makes is causing a world of hurt. I don't know what the exact differences are between the RS690 and RS690M, but I suspect that the M model is missing some of the circuitry contained in the desktop version of the card and if the radeonhd driver is writing to circuitry that doesn't exist in the M model, then that might explain all the problems this chipset has had with the radeonhd module. The chipset does great with fglrx, radeon (not performance wise), and ATI (opensource) modules, but the radeonhd module has given this card fits since the changes in July. Before the July changes, this card did great with the radeonhd builds from the git repository. After the mid-July changes in the repository, the radeonhd driver started blackscreening this box if you recall. I think that same issue that I never could find exactly with bisect is rearing it's ugly head again.
Another big problem is screen artifacts. When you move a window you will have horizontal lines left in the display where the top/bottom of adjacent windows are. See:
Looks like acceleration issues. Does this happen with a minimal xorg.conf like the following?
Section "Device" Identifier "Device[0]" Driver "radeonhd" Option "AccelMethod" "shadowfb" EndSection
Dunno, I'll try it real quick and report back :-)
When we left off, the radeonhd from git was blackscreening my box. Looking at the radeonhd version install (1.3) at least it is running now. (that's a plus).
Yes, indeed.
I am going to pull the latest git version and build it and see how it goes. However, I need to know how I tell 11.2 to load the new driver now --
Install it to /usr/lib(64)/xorg/modules/updates/drivers/ (copying the src/.libs/radeonhd_drv.so is enough). Drivers in this location are preferred over the standard drivers. Don't forget that you need the xorg-x11-server-sdk.rpm for compiling the driver.
since there is no xorg.conf. Do I have to rebuild the kernel initramfs or
Nope. Just the driver.
Thanks for testing
Matthias
-- David C. Rankin, J.D.,P.E. Rankin Law Firm, PLLC 510 Ochiltree Street Nacogdoches, Texas 75961 Telephone: (936) 715-9333 Facsimile: (936) 715-9339 www.rankinlawfirm.com -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org