x86_64 + xine + x264 = Dropped Frames?

[TheeMahn]

wow , once again didn't pay attention, did you build your own kernel (thanks lead)? Lastly the last 64 bit I dropped to the net was Ultimate Edition 1.4 gamers, many a moon ago minus Ultimate Edition 1.7 of course.

drop me the results of:

Code: Select all

dmesg

TheeMahn

I may not have the time now, but will look into to it all.

[maxim]

The kernel that i'm using the default one that came with the 1.7 release. I've not built my own kernel before so I don't even know about that.

I've attached the output of dmesg to this reply. I haven't looked into any possible issues with the codec itself, i'll have to look into that tomorrow, and compare this one against the others that I have.

[maxim]

I'm going to try the 8.04 Release Candidate today some time and see how it works out. I'll let you know my findings.

[maxim]

I booted the system using a Live DVD of Kubuntu 8.04 and it displayed the same symptoms. :-/

[TheeMahn]

The kernel that i'm using the default one that came with the 1.7 release. I've not built my own kernel before so I don't even know about that.

I've attached the output of dmesg to this reply. I haven't looked into any possible issues with the codec itself, i'll have to look into that tomorrow, and compare this one against the others that I have.

holy crap you have 8 cores @ 3.73GHz

I do see you have a bunch of problems with opera, more specifically a plug-in, after I get 1.8 done, I will try and help.

[maxim]

Yeah, it's a 64-Bit ALPHA Version of Opera (Found here: http://snapshot.opera.com/unix/9.50-Alpha-1/), and I think that the flash plugin (or one of the plugins) is only 32bit or something. I really know nothing about it. I just know that I can't live without Opera.

Actually I looked around for a bit and found that Opera didn't like using the other plugins that were install for Firefox. After removing those paths from the plugin path in opera, and installing the x86 version of Adobe Flash the segfault message in dmesg no longer appears, and Flash actually works now.

I also updated to the new and spiffy 9.50 Beta 2 (x86_64) version found here:

http://snapshot.opera.com/unix/snapshot-1929/

I found it through this blog here:

http://my.opera.com/desktopteam/blog/

A few posts down there is a link for the unix version, which isn't displayed on Opera's traditional website. Their site just has i386 versions.

Thanks for provoking me into looking to why my flash wasn't working! Now I have a spiffy new more stable browser, and functioning flash! Yay!

Actually the segmentation fault still appears in dmesg, but it looks to be related to MySpace.com and their crappily written Music player. Pages like YouTube play flash just fine.

[maxim]

I just installed Ultime 1.8 x64 and it's doing the exact same thing. I'm at a loss as to what to do.

[maxim]

I did a little detective work and thinking a little bit and found several threads about seemingly the same problem. It seems this problem occurs globally in Linux, regardless of the platform (Ubuntu, Fedora, etc.).

After looking around I found seval threads on some forums that stated issues similar to mine.

http://ubuntuforums.org/showthread.php?t=692858

http://forum.mymediasystem.org/viewtopic.php?p=10099

http://forum.doom9.org/showthread.php?t=119128

When all of the evidence is put together there are several possible culprits. One is the codec that is being decoded doesn't support the ability to thread it to multiple processors. I can't really check this but i'll looks to see how it plays on another multicore system I have when I get home, it's running Vista unfortunately. Another culprit is the decoder doesn't support threading to multiple processors when decoding this codec (Namely x264). If I understand this properly the decoder is named libavcodec and it is used like a DLL by multiple programs. The last culprit could be xine itself, but that's another can of worms becuase it's just an engine and not a program. I'm familiar with programs but don't know how I would recompile the engine, in addition I would assum that since it's 64bit, it would support SMP by default.

There may be a solution in the "wrapping" of a windows library called CoreAVC for use in Linux. I have no idea how to go about doing that so I'll have to research that over the weekend and see if it will work. There seems to be a project already started and a couple of documents that can be found here:

http://code.google.com/p/coreavc-for-linux/

As I find out more I'll try to remember to keep everyone updated.

[maxim]

Indeed I believe it to be the decoder library that is in use. It only has the ability to use on CPU at any given time. I now have XPx64 on the same hardware and it exibited the same symptoms that it did in Linux, however it was indeed using multiple cores, but playback was still just as bad. Using the CoreAVC codec all cores are being utilized with maximum usage being around 20%, No dropped frames, and no audio syncronization.

I'm going to look into doing that wrapping thing of the CoreAVC in linux over the weekend, I have to reinstall Fedora and Ultimate. Then I'll have the best of all the worlds on one machine. Too bad I have to reboot to get to each one. Oh well. Should be fun.

[maxim]

Alas, I have been unsuccessful in being able to use the CoreAVC within Linux. I attempted to follow the instructions but in Fedora, I could not even get the xine-lib to compile at all. However, in ubuntu I was able to compile it without a hitch, I just had to install some dev packages it was asking for and everything seemed to be OK. Unfortunately the final product was not using the CoreAVC codec at all. I followed the instructions to a T and was unable to get any positive results.