Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
On Sun, Feb 28, 2010 at 4:36 PM, Cyril Brulebois k...@debian.org wrote: Carl Worth cwo...@cworth.org (03/09/2009): Thomas Vaughan wrote: By the way, I don't mind grabbing the relevant source and building for debug output, if that would be of any help. I just need to know how to turn on the relevant output or else where to stick some printf()s. If you're interesting in building things, one thing you might do is to build a recent kernel and use kernel modesetting, (load the i915 module with the option modeset=1 either on the modprobe command line or by putting i915 modeset=1 into /etc/modules). It would be interesting to hear if the behavior is identical with a KMS setup. you should find everything you'd need to test Carl's suggestion in unstable, where KMS is automatically enabled. You can check the details about kernel versions in the second part of this blogpost: http://ikibiki.org/blog/2010/02/28/Where_have_you_been/ Thanks for contacting me. I never followed up on this because the hardware configuration at my office changed. I still do have the machine with the 82945 video adapter, but it now has only a single monitor attached, and I'm using that machine at this point only to serve files. If there be a ticket open somewhere only because I noticed an anomaly that no one else cares about, then one might feel free to close it, even if it be a real anomaly. I don't want to be a source of trouble. :^) The point is that the 82945 can be made to drive a pair of displays, each at 1600x1200, with hardware-accelerated 3D. I got it to work. Unfortunately, after X started, I had to switch to the text VT and back to the X screen in order to get the second screen not to be really sluggish. Anyway, I don't have time to play with it, at least not for the next few weeks. Your message seems to indicate that it should be easy for me to test KMS now that Debian unstable has everything available for free by default. That is an interesting bit of news, and so I might get around to playing with it next month some time. -- Thomas E. Vaughan There are only two kinds of people; those who accept dogma and know it, and those who accept dogma and don't know it. - G.K. Chesterton -- To UNSUBSCRIBE, email to debian-x-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/d9dd60981003010635n5a99a750g2a7fb35bbea50...@mail.gmail.com
Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
Hi Thomas, Carl Worth cwo...@cworth.org (03/09/2009): By the way, I don't mind grabbing the relevant source and building for debug output, if that would be of any help. I just need to know how to turn on the relevant output or else where to stick some printf()s. If you're interesting in building things, one thing you might do is to build a recent kernel and use kernel modesetting, (load the i915 module with the option modeset=1 either on the modprobe command line or by putting i915 modeset=1 into /etc/modules). It would be interesting to hear if the behavior is identical with a KMS setup. you should find everything you'd need to test Carl's suggestion in unstable, where KMS is automatically enabled. You can check the details about kernel versions in the second part of this blogpost: http://ikibiki.org/blog/2010/02/28/Where_have_you_been/ Mraw, KiBi. signature.asc Description: Digital signature
Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
Excerpts from Thomas Vaughan's message of Wed Aug 26 11:57:59 -0700 2009: At 2560x1024 and 3200x1200, the initial condition is for the right monitor to be slow for 2D and 3D. After I do a VT switch via CTRL-ALT-F1 and back (ALT-F7), both 2D and 3D are fully accelerated on the right monitor, at 2560x1024 and at 3200x1200. Hi Thomas, You've got a very mysterious bug. The acceleration code in the driver isn't aware of the two different monitors at all, (it's just seeing coordinates). I discussed the problem with a room full of graphics engineers and mostly got head scratching. But there was one germ of an idea: Perhaps there's some odd RandR transformation being configured for the one display. (For example, imagine the code going through an extra copy and transformation as if rotated.) Something like that might cause the slowdown. And doing the VT switch might get your desktop environment to reconfigure the RandR stuff correctly the second time. That's not much of a theory, but you might poke around to see if the output of xrandr changes from before to after the VT switch. You might also try running with a different desktop environment (or none at all) to see if the buggy behavior changes at all. -Carl -- To UNSUBSCRIBE, email to debian-x-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
Excerpts from Thomas Vaughan's message of Thu Sep 03 16:54:13 -0700 2009: On Thu, Sep 3, 2009 at 12:02 PM, Carl Worthcwo...@cworth.org wrote: You've got a very mysterious bug. Well, at least I have a reliable work-around. Yes, that is nice to have. The output of 'xrandr --verbose' was identical from before to after the VT switch. I diffed the results and got null output. I'll attach one of the output files. I'm not actually all that surprised there, (even if it was some weird RandR thing, the output from xrandr is not too verbose). Moreover, I've done most of my testing by just killing the display manager and starting X with xinit, and then doing 'twm ' at the shell prompt in xterm. OK. So that eliminates the desktop environment as a cause. By the way, do you have access to hardware on which this anomaly can be reproduced? Hmm... are you suggesting that I should perhaps ask for more giant LCD panels. I think I like the way you think. :-) Seriously, though, I will try to setup such an environment and see if I can replicate the behavior. By the way, I don't mind grabbing the relevant source and building for debug output, if that would be of any help. I just need to know how to turn on the relevant output or else where to stick some printf()s. If you're interesting in building things, one thing you might do is to build a recent kernel and use kernel modesetting, (load the i915 module with the option modeset=1 either on the modprobe command line or by putting i915 modeset=1 into /etc/modules). It would be interesting to hear if the behavior is identical with a KMS setup. -Carl -- To UNSUBSCRIBE, email to debian-x-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
On Thu, Sep 3, 2009 at 12:02 PM, Carl Worthcwo...@cworth.org wrote: Hi Thomas, Hello, Carl, Thanks for taking the time to contact me. :^) You've got a very mysterious bug. Well, at least I have a reliable work-around. The acceleration code in the driver isn't aware of the two different monitors at all, (it's just seeing coordinates). I discussed the problem with a room full of graphics engineers and mostly got head scratching. I'm glad that I noticed something that might serve at least as an interesting puzzle. But there was one germ of an idea: Perhaps there's some odd RandR transformation being configured for the one display. (For example, imagine the code going through an extra copy and transformation as if rotated.) Something like that might cause the slowdown. That seems plausible to me. A few months ago I played with rotating one of my monitors, but the performance was so bad that I stopped playing with it. Anyway, the performance problem that I saw with my rotated monitor was of exactly the same kind that I see on the right screen before the VT switch. Not that this proves anything. And doing the VT switch might get your desktop environment to reconfigure the RandR stuff correctly the second time. That's not much of a theory, but you might poke around to see if the output of xrandr changes from before to after the VT switch. The output of 'xrandr --verbose' was identical from before to after the VT switch. I diffed the results and got null output. I'll attach one of the output files. You might also try running with a different desktop environment (or none at all) to see if the buggy behavior changes at all. I recently switched from gnome to kde, but this of course makes no difference. I can see that the right screen is painted slowly even when gdm (or kdm) initializes itself at startup. Moreover, I've done most of my testing by just killing the display manager and starting X with xinit, and then doing 'twm ' at the shell prompt in xterm. The clearest test in that case is just to drag the xterm window from the left screen to the right screen. The twm window manager just draws an outline while dragging, but even this outline is noticeably distorted by motion on the right screen, whereas it is hardly distorted at all on the left. By the way, do you have access to hardware on which this anomaly can be reproduced? It would be nice to know that I am not continually suffering from the same hallucination over and over again. :^) By the way, I don't mind grabbing the relevant source and building for debug output, if that would be of any help. I just need to know how to turn on the relevant output or else where to stick some printf()s. -- Thomas E. Vaughan There are only two kinds of people; those who accept dogma and know it, and those who accept dogma and don't know it. - G.K. Chesterton xrandr-output-before-switch Description: Binary data
Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
On Thu, Aug 20, 2009 at 4:45 PM, Brice Goglinbrice.gog...@ens-lyon.org wrote: Thomas E. Vaughan wrote: Package: xserver-xorg-video-intel Version: 2:2.8.0-2 By the way, the behavior is still the same with 2:2.8.1-1. When the X server first starts up, 2D performance (such as painting the background, dragging a window, etc.) is nice and fast on my left monitor (VGA) but very slow on my right monitor (TMDS-1). I have also now verified that 3D performance is initially slow, too, on the right monitor (whether it be VGA or TMDS-1) but the sluggishness of 3D when 2D is sluggish doesn't seem particularly surprising. By the way, it looks like you're getting a 3200x1200 virtual screen. It may be larger 3D coordinates than what your board supports IIRC. Does it help if you reduce both monitors' resolution to 1024x768 instead of 1600x1200 ? Yes, at 2048x768, the initial condition is fully accelerated for 2D and 3D. At 2560x1024 and 3200x1200, the initial condition is for the right monitor to be slow for 2D and 3D. After I do a VT switch via CTRL-ALT-F1 and back (ALT-F7), both 2D and 3D are fully accelerated on the right monitor, at 2560x1024 and at 3200x1200. For example, after switching to the console and back, I can maximize glxgears on either screen and get about 300 FPS at 3200x1200. Before switching, the right screen gets something abysmal on the order of 10 FPS. -- Thomas E. Vaughan There are only two kinds of people; those who accept dogma and know it, and those who accept dogma and don't know it. - G.K. Chesterton -- To UNSUBSCRIBE, email to debian-x-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
On Thu, Aug 20, 2009 at 4:45 PM, Brice Goglinbrice.gog...@ens-lyon.org wrote: What happens if you swap monitors' positions? Does the slowness follow TMDS? No. That is, when I leave everything unchanged in 'xorg.conf' except for changing RightOf to LeftOf in the Monitor section for what was the right monitor, the initial sluggishness moves to the VGA output. So when the X server fires up, whichever output is designated as being on the right is the one that is slow for window dragging, etc. Again, switching to a console VT and back clears up the problem. By the way, it looks like you're getting a 3200x1200 virtual screen. Yes, I am, and this started working only recently. Aside from the initial sluggishness on the right screen, however, I've seen no problem whatsoever. It may be larger 3D coordinates than what your board supports IIRC. I don't know if this be a good test, but I just ran glxgears, and I'm getting no corruption on either screen, and I get about 300 FPS with glxgears maximized on either screen. Does it help if you reduce both monitors' resolution to 1024x768 instead of 1600x1200 ? Are you wondering whether it helps the initial 2D sluggishness? I don't know yet, but I'll try the experiment at some point and get back to you. By the way, thank you very much for taking the time to investigate this with me. -- Thomas E. Vaughan There are only two kinds of people; those who accept dogma and know it, and those who accept dogma and don't know it. - G.K. Chesterton -- To UNSUBSCRIBE, email to debian-x-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
Package: xserver-xorg-video-intel Version: 2:2.8.0-2 Severity: normal When the X server first starts up, 2D performance (such as painting the background, dragging a window, etc.) is nice and fast on my left monitor (VGA) but very slow on my right monitor (TMDS-1). I noticed by accident that this problem is worked around by manually doing a VT switch to the console (CTRL-ALT-F1) and then doing a switch back (ALT-F7). After this, performance is very nice on both monitors. -- Package-specific info: /var/lib/x11/X.roster does not exist. /var/lib/x11/X.md5sum does not exist. X server symlink status: lrwxrwxrwx 1 root root 13 2009-07-02 09:53 /etc/X11/X - /usr/bin/Xorg -rwxr-xr-x 1 root root 1867584 2009-08-01 01:45 /usr/bin/Xorg /var/lib/x11/xorg.conf.roster does not exist. VGA-compatible devices on PCI bus: 00:02.0 VGA compatible controller: Intel Corporation 82945G/GZ Integrated Graphics Controller (rev 02) /var/lib/x11/xorg.conf.md5sum does not exist. Xorg X server configuration file status: -rw-r--r-- 1 root root 375 2009-08-20 12:40 /etc/X11/xorg.conf Contents of /etc/X11/xorg.conf: Section Monitor Identifier Left EndSection Section Monitor Identifier Right Option RightOf Left EndSection Section Device Identifier 82945G/GZ Driver intel Option monitor-TMDS-1 Right Option monitor-VGA Left EndSection # Allow KDE 4.3.0 to run without corrupting the screen. Section Extensions Option Composite Disable EndSection Xorg X server log files on system: -rw-r--r-- 1 root root 31506 2009-07-02 09:54 /var/log/Xorg.1.log -rw-r--r-- 1 root root 37778 2009-08-20 12:43 /var/log/Xorg.0.log Contents of most recent Xorg X server log file /var/log/Xorg.0.log: X.Org X Server 1.6.3 Release Date: 2009-7-31 X Protocol Version 11, Revision 0 Build Operating System: Linux 2.6.30.2-dsa-amd64 x86_64 Debian Current Operating System: Linux hypostasis 2.6.30-1-amd64 #1 SMP Sat Aug 15 18:09:19 UTC 2009 x86_64 Build Date: 01 August 2009 07:44:36AM xorg-server 2:1.6.3-1 (buildd@) Before reporting problems, check http://wiki.x.org to make sure that you have the latest version. Markers: (--) probed, (**) from config file, (==) default setting, (++) from command line, (!!) notice, (II) informational, (WW) warning, (EE) error, (NI) not implemented, (??) unknown. (==) Log file: /var/log/Xorg.0.log, Time: Thu Aug 20 12:41:24 2009 (==) Using config file: /etc/X11/xorg.conf (==) No Layout section. Using the first Screen section. (==) No screen section available. Using defaults. (**) |--Screen Default Screen Section (0) (**) | |--Monitor default monitor (==) No device specified for screen Default Screen Section. Using the first device section listed. (**) | |--Device 82945G/GZ (==) No monitor specified for screen Default Screen Section. Using a default monitor configuration. (==) Automatically adding devices (==) Automatically enabling devices (==) FontPath set to: /usr/share/fonts/X11/misc, /usr/share/fonts/X11/cyrillic, /usr/share/fonts/X11/100dpi/:unscaled, /usr/share/fonts/X11/75dpi/:unscaled, /usr/share/fonts/X11/Type1, /usr/share/fonts/X11/100dpi, /usr/share/fonts/X11/75dpi, /var/lib/defoma/x-ttcidfont-conf.d/dirs/TrueType, built-ins (==) ModulePath set to /usr/lib/xorg/modules (**) Extension Composite is disabled (II) Cannot locate a core pointer device. (II) Cannot locate a core keyboard device. (II) The server relies on HAL to provide the list of input devices. If no devices become available, reconfigure HAL or disable AllowEmptyInput. (II) Loader magic: 0x3740 (II) Module ABI versions: X.Org ANSI C Emulation: 0.4 X.Org Video Driver: 5.0 X.Org XInput driver : 4.0 X.Org Server Extension : 2.0 (II) Loader running on linux (++) using VT number 7 (--) PCI:*(0:0:2:0) 8086:2772:1028:01ad Intel Corporation 82945G/GZ Integrated Graphics Controller rev 2, Mem @ 0xfeb0/524288, 0xe000/268435456, 0xfeac/262144, I/O @ 0xe898/8 (II) Open ACPI successful (/var/run/acpid.socket) (II) System resource ranges: [0] -1 0 0x - 0x (0x1) MX[B] [1] -1 0 0x000f - 0x000f (0x1) MX[B] [2] -1 0 0x000c - 0x000e (0x3) MX[B] [3] -1 0 0x - 0x0009 (0xa) MX[B] [4] -1 0 0x - 0x (0x1) IX[B] [5] -1 0 0x - 0x (0x1) IX[B] (II) LoadModule: extmod (II) Loading /usr/lib/xorg/modules/extensions//libextmod.so (II) Module extmod: vendor=X.Org Foundation compiled for 1.6.3, module version = 1.0.0 Module class: X.Org Server Extension ABI class: X.Org Server Extension, version 2.0 (II) Loading extension SELinux (II) Loading extension MIT-SCREEN-SAVER (II) Loading extension XFree86-VidModeExtension (II) Loading extension
Bug#542701: xserver-xorg-video-intel: Very slow 2D performance on one screen of dual display until after switch to console VT and back.
Thomas E. Vaughan wrote: Package: xserver-xorg-video-intel Version: 2:2.8.0-2 Severity: normal When the X server first starts up, 2D performance (such as painting the background, dragging a window, etc.) is nice and fast on my left monitor (VGA) but very slow on my right monitor (TMDS-1). What happens if you swap monitors' positions? Does the slowness follow TMDS? By the way, it looks like you're getting a 3200x1200 virtual screen. It may be larger 3D coordinates than what your board supports IIRC. Does it help if you reduce both monitors' resolution to 1024x768 instead of 1600x1200 ? Brice -- To UNSUBSCRIBE, email to debian-x-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org