View previous topic :: View next topic |
Author |
Message |
WD-40 n00b
Joined: 22 Sep 2004 Posts: 53
|
Posted: Wed Sep 26, 2012 3:48 am Post subject: Disabling "Composite" and "RENDER" - Is |
|
|
I'm attempting to troubleshoot some performance issues on an Atom / Intel 945GME netbook, and one of the steps I tried was to disable the "Composite" and "RENDER" extensions for Xorg. My goal is to improve full-screen video playback, so I believe that neither of these extensions will be of any use to me.
To my surprise, they were automatically re-enabled!
tail end of xorg.conf:
Code: | Section "Extensions"
Option "Composite" "Disable"
Option "RENDER" "Disable"
EndSection
|
clips of resulting Xorg.0.log:
Code: | [ 607.372] (**) Extension "Composite" is disabled
[ 607.372] (**) Extension "RENDER" is disabled
... and then a bit later ...
[ 607.528] (II) Initializing built-in extension Generic Event Extension
[ 607.528] (II) Initializing built-in extension SHAPE
[ 607.528] (II) Initializing built-in extension MIT-SHM
[ 607.528] (II) Initializing built-in extension XInputExtension
[ 607.528] (II) Initializing built-in extension XTEST
[ 607.528] (II) Initializing built-in extension BIG-REQUESTS
[ 607.528] (II) Initializing built-in extension SYNC
[ 607.528] (II) Initializing built-in extension XKEYBOARD
[ 607.528] (II) Initializing built-in extension XC-MISC
[ 607.528] (II) Initializing built-in extension XINERAMA
[ 607.528] (II) Initializing built-in extension XFIXES
[ 607.528] (II) Initializing built-in extension RENDER
[ 607.528] (II) Initializing built-in extension RANDR
[ 607.528] (II) Initializing built-in extension COMPOSITE
[ 607.528] (II) Initializing built-in extension DAMAGE
|
So it is clear that Xorg read the config file and processed my extension disablings... however, the Intel driver went ahead and enabled them anyway?
I then attempted to disable them more directly, in the "Device" section along with the other Intel parameters, with the line:
Code: | Option "RENDER" "False" |
... but the driver doesn't like that:
Code: | [ 607.527] (WW) intel(0): Option "RENDER" is not used |
So my question after all of this:
Is it still possible to disable the "Composite" and "RENDER" extensions in Xorg? And if so, how do you do it?
Thanks for the help... |
|
Back to top |
|
|
Gusar Advocate
Joined: 09 Apr 2005 Posts: 2665 Location: Slovenia
|
Posted: Wed Sep 26, 2012 5:25 am Post subject: |
|
|
It's probably possible, but you won't gain better video playback by doing that. Instead of trying to use what you think is a solution, you should instead describe the original problem, because the actual solution to it might not be what you think.
The 945GME has no hardware decoder, so all it provides it accelerating presentation via the Xv extension. This requires working DRI to use textured video. Then there's the hardware overlay, modern graphic cards don't have that anymore, but the 945GME is not modern .
So, do you have DRI (3d accel) working? Is your player configured to use Xv? |
|
Back to top |
|
|
WD-40 n00b
Joined: 22 Sep 2004 Posts: 53
|
Posted: Sat Sep 29, 2012 12:09 am Post subject: |
|
|
Gusar wrote: | It's probably possible, but you won't gain better video playback by doing that. Instead of trying to use what you think is a solution, you should instead describe the original problem, because the actual solution to it might not be what you think. |
Sure. The problem is that I get a periodic slowdown in the video framerate, very briefly, and perhaps once per second. The video is being played with the FFmpeg-based player in MythTV, and is using a Broadcom 'Crystal HD' hardware video decoder. The output renderer is 'xv-blit', DRI is enabled according to glxinfo, and all post-processing (advanced deinterlacing, etc) is disabled. The video content is 1080i broadcast ATSC MPEG-2, and the display is a VGA-connected LCD running at 1920x1080 at 16-bit color.
Power management is disabled in the kernel, as the Intel drivers have trouble with video performance when the CPU throttles down. With the CrystalHD decoder, CPU usage is between 40-60% for 1080i content, with a bit over half of that being system usage. 'Wait' CPU is extremely low, < 1%, and iostat shows only a very modest amount of disk activity. The recordings being played are on a freshly-formatted JFS partition on a hard drive, and hdparm verifies that the disk has 10x+ the throughput required. These are pre-recorded shows, so the system isn't busy with the tuner or with bi-directional I/O.
So my logic is that the disk I/O checks out, the CPU isn't pegged (40%+ free), and the CrystalHD is working (video is unwatchable with 100% cpu when it is disabled). So the two possibilities remaining then are that either the CrystalHD isn't being fed properly, or the graphics system can't push the resulting pixels out fast enough.
... which brings me to the question for this thread. Alpha-blending, an extra off-screen buffering, and the compositing are all extra steps in that final stage of pushing pixels out the pipe - hence my (intended) test to disable them.
If you see anything else I can check though, please let me know - I appreciate any help you might have with this. |
|
Back to top |
|
|
Gusar Advocate
Joined: 09 Apr 2005 Posts: 2665 Location: Slovenia
|
Posted: Sat Sep 29, 2012 5:23 pm Post subject: |
|
|
WD-40 wrote: | Power management is disabled in the kernel, as the Intel drivers have trouble with video performance when the CPU throttles down. |
Err, what? That's the first time I hear about anything like that. Do you have any links that talk about this?
WD-40 wrote: | So the two possibilities remaining then are that either the CrystalHD isn't being fed properly, or the graphics system can't push the resulting pixels out fast enough. |
It's for sure CrystalHD. That thing is very delicate, driving it properly is an art form. Which driver are you using, the one in the kernel, or the one provided by Broadcom? And which model exactly do you have? That's important info, as they behave quite differently.
WD-40 wrote: | Alpha-blending, an extra off-screen buffering, and the compositing are all extra steps in that final stage of pushing pixels out the pipe |
Are you running a compositor? If not, all that stuff is irrelevant, as it will only be done when a compositor is running.
BTW, that you use CrystalHD was completely absent from your fist post. You didn't think it was important info that you should share? That's why I said to describe the problem, not what you think the solution is. |
|
Back to top |
|
|
WD-40 n00b
Joined: 22 Sep 2004 Posts: 53
|
Posted: Sun Sep 30, 2012 1:30 am Post subject: |
|
|
Gusar wrote: | WD-40 wrote: | Power management is disabled in the kernel, as the Intel drivers have trouble with video performance when the CPU throttles down. |
Err, what? That's the first time I hear about anything like that. Do you have any links that talk about this? |
Yep, here's the bug track for it:
https://bugs.freedesktop.org/show_bug.cgi?id=30364
Gusar wrote: | WD-40 wrote: | So the two possibilities remaining then are that either the CrystalHD isn't being fed properly, or the graphics system can't push the resulting pixels out fast enough. |
It's for sure CrystalHD. That thing is very delicate, driving it properly is an art form. Which driver are you using, the one in the kernel, or the one provided by Broadcom? And which model exactly do you have? That's important info, as they behave quite differently. |
I have the BCM970015, which should be more tolerant than the -12, from what I have read.
I started with the kernel staging driver, but never got it to load fully (no /dev/crystalhd device created, and modprobe & dmesg output didn't indicate any failures...). So I switched to Jarod's GIT repository: http://git.linuxtv.org/jarod/crystalhd.git
Gusar wrote: | WD-40 wrote: | Alpha-blending, an extra off-screen buffering, and the compositing are all extra steps in that final stage of pushing pixels out the pipe |
Are you running a compositor? If not, all that stuff is irrelevant, as it will only be done when a compositor is running. |
Nope, not that I know of. Just Window Maker as a window manager.
Gusar wrote: | BTW, that you use CrystalHD was completely absent from your fist post. You didn't think it was important info that you should share? That's why I said to describe the problem, not what you think the solution is. |
To be fair, I did describe the problem. The problem I was troubleshooting here was that the Intel graphics driver isn't following the configuration in the xorg.conf file, and is giving contradictory status output in the log file.
The video playback issue is separate, and I definitely appreciate your help in resolving that. But regardless of whether I solve that, the contradictory Xorg output is still a problem of its own. |
|
Back to top |
|
|
Gusar Advocate
Joined: 09 Apr 2005 Posts: 2665 Location: Slovenia
|
Posted: Sun Sep 30, 2012 8:55 am Post subject: |
|
|
That's about 3D performance, not about video. And there's a simple workaround mentioned (granted, it might cause tearing) - playing with vsync settings. I don't use any such workarounds and video playback works fine on my netbook. It's software decoded though, no CrystalHD here. |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|