Created attachment 125576 [details] Xorg.0.log VIVID - working After updating the Ubuntu HWE stack in 14.04 to Xenial (kernel 4.4 and xserver-xorg-video-radeon-lts-xenial), system boots up fine and responsively, but screen is laggy; there is a tremendous performance regression in comparison to xorg-lts-vivid. Lots of video tearing. System is an older Pentium laptop (Packard Bell EasyNote) with a Radeon Xpress 200M chip (r300 driver). Video acceleration was working with DRI2 under the Vivid stack (kernel 3.19-64-lowlatency + xserver-xorg-video-radeon-lts-vivid). I also tried upgrading to the Wily kernel (4.2.0-lowlatency), and video acceleration works fine. I reported this in launchpad previously, and it was dismissed as "invalid" since Wily is EOL. The problem persists in the Xenial stack, so I'm reporting it here. Attached are the respective Xorg.0 logs.
Created attachment 125577 [details] Xorg.0.log XENIAL - regression
Here is the link to the original report in Launchpad: https://bugs.launchpad.net/ubuntu/+source/xorg/+bug/1598593
I've just upgraded the kernel to 4.4.0-34 (lowlatency). (I've also corrected the Xorg versions; as Ubuntu user, I've fallen into the bad habit of naming the package versions by the distribution nickname. Sorry.) Video acceleration works just fine combined with the X server for radeon, version 7.5, using the r300 driver and DRI2. Problems begin when using X servers 7.6 and 7.7. Tearing ensues, regardless of the presence of a xorg.conf config file to enable/disable options to improve framerate. Enabling DRI3 only worsens the performance regression, and accentuates tearing. What I've identified is: both xorg versions support DRI3, and show a performance regression even if only DRI2 is active. Xorg 7.5 works great for the r300 driver on the Radeon Xpress 200M onboard chip. Here are 2 videos: the first is using Xorg 7.5, YouTube fullscreen on Chromium. The second video is the same, under 7.7 (same problem as 7.6). 7.5: https://youtu.be/Ykhp2L_pf7Y 7.7: https://youtu.be/p7FvXTzlIa0 (ditto for 7.6) The regression is evident.
Comment on attachment 125576 [details] Xorg.0.log VIVID - working Xorg version is 7.5
Comment on attachment 125577 [details] Xorg.0.log XENIAL - regression Xorg version is 7.7
So the problem also occurs with xf86-video-ati 7.6.0 or newer on Vivid? If so, can you bisect upstream xf86-video-ati Git?
Hello Michel, I've tried doing that, but it breaks the installation, Ubuntu refuses ("broken packages"). How could one do that properly?
All I lnow is that the last working version is 7.5.0-1, and the first regression is observed in version 7.5.0+git20150819. If there's more information needed, please let me know. Please bear in mind that this laptop is a production machine with limited capabilities to run a VM for testing purposes.
(I misinterpreted "bisect" as "test" the first time, sorry)
(In reply to Dennis Mayr from comment #7) > I've tried doing that, but it breaks the installation, Ubuntu refuses > ("broken packages"). I guess you'd have to compile the newer source packages against the Vivid system. (In reply to Dennis Mayr from comment #8) > All I lnow is that the last working version is 7.5.0-1, and the first > regression is observed in version 7.5.0+git20150819. There were a lot of changes between those two, so without a bisection it'll be hard if not impossible to figure out what caused the problem.
Hello Michel, I'll ask someone to help me out doing this, to pinpoint the problematic build. Cheers!
I was finally trying to bisect, but this machine has run out of drive space while compiling the drivers (old machine, tight budget), and there's not enough time off-production to keep trying in an efficient way, so I'll have to ask someone else to help me out building the packages.
I own an Xpress 200m chip as well and can confirm this issue. I will also try to bisect, but I run into a dependency problem: (try to compile 7.5.0) > ./configure: line 18738: syntax error near unexpected token `RANDR,' > ./configure: line 18738: `XORG_DRIVER_CHECK_EXT(RANDR, randrproto)' My system is openSUSE Tumbleweed (=rolling release) and I have install several -devel packages. Could you tell me which devel package is still missing? (or which files are missing?)
(In reply to Dominik Kopp from comment #13) > > ./configure: line 18738: syntax error near unexpected token `RANDR,' > > ./configure: line 18738: `XORG_DRIVER_CHECK_EXT(RANDR, randrproto)' The /usr/share/aclocal/xorg-macros.m4 file is missing. Re-run autogen.sh after installing the package providing that.
Thanks, but that file was already there and re-run autorun.sh results in the same the error again. I've made also a new try in a fresh directory, but the error message remains. Any other ideas?
My bad, it's /usr/share/aclocal/xorg-server.m4 .
It's quite hard to make it compile, if the version is older than b6d871b (I assume the x-server 1.18 is to new for older xf86-video-ati versions.) I've made a simple "make", "sudo make install", reboot. But I could find a working version, yet. Therefore I started another approach: I reinstalled the laptop with a working openSUSE 13.2 - kernel 3.16.7 - x-server 1.16 - xf86-video-ati 7.4.0 Then I updated it *only with 7.7.0 / 1.18 - working Then I updated it to kernel 4.7.2 - regression (with and without the newer x-server/xf86-video-ati version) I conclude: (at least for me) The regression is not in the xorg, but in the kernel. I need to bisect the kernel (somewhere between 3.16.7 and 4.7.2) PS: An additional argument for kernel bisecting is the running fan at full speed. PPS: Note to myself: file /usr/share/aclocal/xorg-server.m4 is in package util-macros-devel
I'm going to test the xserver-xorg-video-ati 7.7 test package that Timo Aaltonen has built for Ubuntu, this time along with kernel 3.16 and 3.19 as well, to check Dominik's report.
Ok, I just finished testing the following: -Xserver: 1.18 -xserver-xorg-video-ati: 7.7 On kernels: -3.13.0-95 -3.16.0-77 -3.19.0-68 -4.2.0-42 -4.4.0-36 Video lags on all kernels. When using: -Xserver: 1.16 -xserver-xorg-video-ati: 7.5 On the above kernels, video and system GUI (Gnome/GTK/KDE) run smoothly, at least in my case (Ubuntu 14.04). Just taking a wild guess (not my field): could this hypothetically be remotely related to a DRI module?
(In reply to Dominik Kopp from comment #17) > I conclude: (at least for me) > The regression is not in the xorg, but in the kernel. Since this doesn't seem to be the case for Dennis, we should assume for now that you guys are running into different problems. Dominik, please file your own report for product DRI, component DRM/Radeon.
(In reply to Dennis Mayr from comment #19) > Ok, I just finished testing the following: > -Xserver: 1.18 > -xserver-xorg-video-ati: 7.7 [...] > When using: > -Xserver: 1.16 > -xserver-xorg-video-ati: 7.5 There are still two variables between those tests. Can you test xserver-xorg-video-ati 7.7 with xserver 1.16? > Just taking a wild guess (not my field): could this hypothetically be > remotely related to a DRI module? Doesn't look like it from the log files you attached.
Hello Michel, Ubuntu won't let me because of broken dependencies; it will only allow to install the full stack. However, Timo Aaltonen from Ubuntu helped me, bisecting versions between 7.5.0-1 and 7.5.0+git20150819, for me to test on my computer via PPA. What we found, however, is that the last bisection ended up being version 7.5.0, and there still was a lot of tearing, so the culprit may be something else. Mesa is one of the suspects. At this point, I'd give up, don't want this to become a waste of developer's time. Maybe I'll just stick to the older stack, which works perfectly. If anybody else wants to pursue the bughunt, be aware that there may be other parts of the stack thay may cause the problem. Thanks for the help and involvement, guys!
-- GitLab Migration Automatic Message -- This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity. You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/xorg/driver/xf86-video-ati/issues/162.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.