Updating xorg radeon driver

Help with using MiniMyth's build system

Moderator: Pablo

Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Updating xorg radeon driver

Post by Glidos »

Minimyth by bad luck has picked a version of the radeon driver from a period where some interlaced modes were broken. I'm assured it's all fixed again on ati git master. I was thinking of having a go at building minimyth with the update. Any hints on doing this - quick overview or the like? I'd be using SuSE 11.0 as the host for the build.

I'm not sure that the latest radeon driver works with the version of xorg that minimyth uses, so it may not be a goer.
Pablo
Site Admin
Posts: 4182
Joined: Tue Dec 14, 2004 2:13 am
Location: La Jolla
Contact:

Re: Updating xorg radeon driver

Post by Pablo »

Glidos wrote:Minimyth by bad luck has picked a version of the radeon driver from a period where some interlaced modes were broken. I'm assured it's all fixed again on ati git master. I was thinking of having a go at building minimyth with the update. Any hints on doing this - quick overview or the like? I'd be using SuSE 11.0 as the host for the build.
Beyond the MiniMyth build instructions, I do not have any advice.
I'm not sure that the latest radeon driver works with the version of xorg that minimyth uses, so it may not be a goer.
I switched from the latest release version to the latest git version. The latest git version of xf86-video-ati driver builds with the rest of MiniMyth's Xorg packages. However, I test it.

The latest test build (0.21.0-62b4) contains the latest git version of xf86-video-ati. Could you let me know whether or not it fixes your problem with interlaced modes?
MiniMyth running on an Acer ApireRevo 3610 and a Zotac ZBOX-ID80-U. Find out more at my MythTV page.
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Wow, thanks Pablo. That's really good of you. I'll try it tonight. Can't wait!
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Yes, that fixed it. With 62b4, I can get my Radeon 9000 to generate a PAL compatible signal and run it into my TVs Scart.

I still have problems. I can't get very smooth playback as I was with v4.2.0 of the driver, but then that was with MythTV installed under SuSE, so there are a lot of variables to consider. I'll try the latest radeon driver also under SuSE to see if that's the issue.

I also tried 62b4 on my live via-based system. Works great.

Thanks again Pablo.
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Seems I still have problems. I've reinstalled SuSE 11.0 onto my test system that has the Radeon 9000, and updated the ati/radeon driver to 6.9.0 to match 62b4 of Minimyth. I can now run the same hardware with both SuSE and Minimyth with same radeon driver version, and both able to drive my TV's Scart. Provided I use XVMC, SuSE gives me perfectly smooth playback, but I don't seem to be able to get the same from Minimyth. It's not jitter. It's something more subtle. Perhaps I'm getting a consistent half frame rate. I get much the same lack of smoothness out of SuSE if I don't use XVMC.

The problem is that Minimyth is what I want to use really. I'm a bit lost as to how to track down the cause. I suppose I could try to build a really cut down version of SuSE that can boot using nfs, but I'd much rather stick with the simplicity of Minimyth.
Pablo
Site Admin
Posts: 4182
Joined: Tue Dec 14, 2004 2:13 am
Location: La Jolla
Contact:

Post by Pablo »

Many things other than the driver are likely different between SuSE 11.0 and MiniMyth.

Compiler version and options can impact performance. What variation of SuSE are you running (i486, i586, i686, x86_64, etc) and what flags were used when it was compiled?

User permissions can impact performance. When you run mythfrontend on SuSE are you running as a user with restricted access or as root?

Kernel version and configuration can impact performance. Are there differences in how the SuSE and MiniMyth kernels have been configured to generate interrupts and do preemption?

Differences in Xorg configuration and package versions can impact performance. Are there differences in the xorg.conf files for each distribution? Are there differences in the /var/log/Xorg* log files for each distribution?
MiniMyth running on an Acer ApireRevo 3610 and a Zotac ZBOX-ID80-U. Find out more at my MythTV page.
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Yeah, but what seems odd to me is that I have the problem only with this hardware: Minimyth is perfectly smooth on my VIA board... and obviously with most other systems, otherwise people would be reporting problems. It's just on this other system that it's not running right, and it should be faster than the VIA board as far as I can see (A8V mobo, A64 3200+ CPU and Radeon 9000). Oh, it is using DDR1, whereas the VIA board is DDR2.

If SuSE wasn't smooth then I'd suspect the harware, but SuSE is fine, with XVMC at least. Even if I can find differences between SuSE and Minimyth, it's still odd for Minimyth to run well on the low-power VIA board, and not on a 3200+. I was hoping that the oddness of that might point to what it might be.

Re things you mentioned:
I don't know the comiler version. On SuSE I'm using all precompiled packages.

It's built for i586. I don't know what flags; how would I find out?

I've been running as a restricted user, to match Minimyth, and because you told me that running as root can cause problems. I think I've also had it work ok as root too. I could check.

Kernel is 2.6.25.5-1.1-pae It's the stock kernel. I don't know about the configuration for generating interrupts and preemption.

Xorg: I have both xorg.conf files and both Xorg.0.log files. In the SuSE one there looked to be something about dri failing. Could that possibly lead to an advantage? I could post these files, or the diff.

If it were the other way around, I wouldn't find it odd because I assume you've picked all the best options for MythTV in Minimyth., so it wouldn't be at all odd if Minimyth outperformed SuSE, but the other way around is strange. Surely it has to be something very specific to this hardware.

Oh, one other possibly big diefference. I have SuSE on a HDD, whereas I'm using Minimyth via NFS. Is that a possible problem?
Pablo
Site Admin
Posts: 4182
Joined: Tue Dec 14, 2004 2:13 am
Location: La Jolla
Contact:

Post by Pablo »

If you pm me your /etc/X11/xorg.conf and /var/log/Xorg.*.log files for both, then will take a look and see whether or not I can find anything.

It seems unlikely that the HDD/NFS difference would matter. After all, once it is playing, everything it needs should be in memory.
MiniMyth running on an Acer ApireRevo 3610 and a Zotac ZBOX-ID80-U. Find out more at my MythTV page.
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Brilliant! Thanks Pablo. I'll send them after this post.

Possibly you wont need to look at them because I've found something in the Xorg logs.

From SuSE

Code: Select all

(==) RADEON(0): Using AGP 4x
(II) RADEON(0): [agp] Mode 0x1f000207 [AGP 0x1106/0x0282; Card 0x1002/0x4966]
(II) RADEON(0): [agp] 32768 kB allocated with handle 0x00000001
(II) RADEON(0): [agp] ring handle = 0xe0000000
(II) RADEON(0): [agp] Ring mapped at 0xb3907000
(II) RADEON(0): [agp] ring read ptr handle = 0xe0101000
(II) RADEON(0): [agp] Ring read ptr mapped at 0xb7fd2000
(II) RADEON(0): [agp] vertex/indirect buffers handle = 0xe0102000
(II) RADEON(0): [agp] Vertex/indirect buffers mapped at 0xb3707000
(II) RADEON(0): [agp] GART texture map handle = 0xe0302000
(II) RADEON(0): [agp] GART Texture map mapped at 0xb1a87000
(II) RADEON(0): [drm] register handle = 0xfbf00000
(II) RADEON(0): [dri] Visual configs initialized
From Minimyth:

Code: Select all

(WW) RADEON(0): [agp] AGP not available
(EE) RADEON(0): [agp] AGP failed to initialize. Disabling the DRI.
(II) RADEON(0): [agp] You may want to make sure the agpgart kernel module
is loaded before the radeon kernel module.
(II) RADEON(0): RADEONDRICloseScreen
(II) RADEON(0): [drm] removed 1 reserved context for kernel
(II) RADEON(0): [drm] unmapping 8192 bytes of SAREA 0xf8aa8000 at 0xb7bc4000
(II) RADEON(0): [drm] Closed DRM master.
(II) RADEON(0): Initializing fb layer
Could that be it?
Pablo
Site Admin
Posts: 4182
Joined: Tue Dec 14, 2004 2:13 am
Location: La Jolla
Contact:

Post by Pablo »

Yes that could very easily be the problem. Support for 0x1106/0x0282 is not in MiniMyth. I have added it and uploaded a new test build: 0.21.0-62b7. Also, this test build contains the lastest GIT version of xf86-video-ati.

Looking at the /var/log/Xorg.*.log files, the only other difference that I could see was the difference in GARTSize. It is set to 32 by the /etc/X11/xorg.conf file for SuSE. Therefore, while the MiniMyth uses the default value of 8MB, SuSE uses 32MB. While I do not think that this should be an issue, if having AGP support does not solve the problem, then you might try adding the GARTSize option to the /etc/X11/xorg.conf file for MiniMyth.
MiniMyth running on an Acer ApireRevo 3610 and a Zotac ZBOX-ID80-U. Find out more at my MythTV page.
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Great! Will try that out tonight. Thanks again Pablo.
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Thanks for trying, but no, still not running right. It did make the expected difference in Xorg.0.log, but no difference to performance.

I'll pm the log in case that's useful.
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Here's an evil idea, could I make my own minimyth-like version of SuSE. Here's what I'm thinking:

1) Put together the intended hardware, initially including a HDD with SuSE and MythTV on it.

2) Run it for a week, using all the various features and formats, and then make a list of all the files that have last-accessed times within that week.
I'm hoping those are the only files I need.

3) On my server, take an nfs minimyth distro, and replace rootfs-ro by the required files as determined by the previous step.

4) Replace the files in the minimyth top-level bin, lib, sbin etc. by the equivalent SuSE files.

5) Replace the minimyth kernel by the SuSE kernel.

6) Remove the harddrive from the client, network boot as I do for minimyth.

Any chance something like that would work?
Pablo
Site Admin
Posts: 4182
Joined: Tue Dec 14, 2004 2:13 am
Location: La Jolla
Contact:

Post by Pablo »

Glidos wrote:Here's an evil idea, could I make my own minimyth-like version of SuSE. Here's what I'm thinking:

1) Put together the intended hardware, initially including a HDD with SuSE and MythTV on it.

2) Run it for a week, using all the various features and formats, and then make a list of all the files that have last-accessed times within that week.
I'm hoping those are the only files I need.

3) On my server, take an nfs minimyth distro, and replace rootfs-ro by the required files as determined by the previous step.

4) Replace the files in the minimyth top-level bin, lib, sbin etc. by the equivalent SuSE files.

5) Replace the minimyth kernel by the SuSE kernel.

6) Remove the harddrive from the client, network boot as I do for minimyth.

Any chance something like that would work?
This would be difficult but it may be possible. However, whether or not it is possible depends on what characteristics of MiniMyth you are trying to preserve. So, what about MiniMyth are you trying to preserve?
MiniMyth running on an Acer ApireRevo 3610 and a Zotac ZBOX-ID80-U. Find out more at my MythTV page.
Glidos
Contributor
Posts: 164
Joined: Tue Sep 02, 2008 8:50 am

Post by Glidos »

Main desire is booting from a read only nfs share, and to keep it small if possible. I might yet discover what's upsetting minimyth, but useful to have another plan failing that.

... so I guess none of the main features, such as running on lots of hardware, with simple config. I wouldn't be looking to run SuSE on any other hardware.
Post Reply