Log in

View Full Version : Virtual dub crash with Koepi Xvid


David caid
9th September 2002, 13:33
Hello,

I post this thread in the Xvid forum and not in the virtualdub because,
my problem seems to be related only to XVID codecs.

I have a crash with VirtualDub when I go to video processing and try to configure XVID.
When I click on the configure button, Virtual dub crashes.
The crash info is in the attached file (crashinfo.txt).

This problem occurs with Koepi versions since the 30th of June.
Version tested :
30062002-1
04082002-1
18082002-1
30082002-1
04092002-1

The lastest version working is the 28062002-1
The Nic version is working (but the last one is from the 28062002).

I didn't see any major changes in the installation except the xvid.dll file (and also the alt_curve.gif it should not influence Virtual dub).
I think the version of 30th of June is the first to introduce EPSZ(^2) Motion Estimation activated. Is this related ?

So the important thing is my configuration.
I have a AMD K6-2 400 MHz with WIN98SE. Could it be because of the compilation of the sources ? Version of 04092002 was compiled to avoid processor problem, no ?
Is that OS related ?
As I reinstalled my computer, I have a clean system.
I only have Xvid codec, ffdshow, oggDS (version 0.991), ogg subtitle filter, Virtualdub 1.4.10

Does anybody encounter the same problem ?

In avance, thanks.

David

vlad59
9th September 2002, 13:49
IIRC, Koepi's build are compiled with intel compilers with special optimizations -> so they won't work with PII or K6-II.
Koepi have forgot to use those optimization on the first version 04092002 so that explain your results.

It was explained better in another thread (make a search to find it)

Hope this helps

rui
9th September 2002, 13:52
If i am not correct, the latest builds from Koepi are compiled for performance, and have some problems with AMD K6*
I think that, when you say "...was compiled to avoid processor problem, no", you probably read a thread where Koepi says that he compiled to be processor compatible, but he was meaning the statsreader.

Koepi, please correct me if i am wrong.

EDIt: sorry to answer something that Vlad already had. He was too fast for me ;)

David caid
9th September 2002, 14:18
Hello vlad59,

I found this thread where it said the Xvid 29082002 is working with K6-2 machnes (A version I didn't download from Koepi site :( )
http://forum.doom9.org/showthread.php?threadid=32294&perpage=20&highlight=intel&pagenumber=1

Does it means that now Koepi is using special optimizations definitively so that I cannot use my so powerfull :) K6 for encoding with the lastest
version of Xvid ?

David

vlad59
9th September 2002, 14:27
You'll have to wait for Koepi's answer !!!
I have no idea about it, but I understand your feeling I just dropped my old K6-II 400 to buy a Duron 1.2Ghz.

David caid
9th September 2002, 15:09
Well well well,

buying a new machine is not on the way right now, therefore, I'm proud of my 3 fps encoding :) Well, the encoding time is not important for playback, and I can still reach 25 fps while reading my Xvid.

No seriously, I hope, I can still use the new version of Xvid. as it is really a great codec.

Koepi, I really appreciate what you are doing for us for Xvid codec and also the oggmux (it was a great improvement not to use graphedit anymore). I hope you can find an easy solution for this problem. If you need to make a special compilation for K6-2 and P2 owner and if it is bothering you, I would understand.

Thanks Vlad59 and rui for your answers

David

Marc FD
9th September 2002, 15:49
I would use uManiacs builds.
They are not optimised, but you won't see the difference on a K6-II.
I have a K6-II too, and i would do so if i wanted to encode XviD with it.
Or do you absolutly want EPZS^2 (who are not _better_) ?

Koepi
9th September 2002, 17:59
Your info about EPZS(^2) is misleading.

It's not better in any case. It's not worse in any case.
EPZS assumes more motion than plain PMVfast and thus can help action based movies in particular as the motion vectors found with it need less bits on motion compensation/error compensation.
(That's the theory behind it at least).
The implementation isn't 100% in XviD, but close by it. (Still it's clearly experimental)

My binaries are compiled with optimizations that makes them crash on k6/k6-2/pII and earlier. I don't intnd to build 2 binaries, that'd be overkill, sorry (my bandwidth is abused highly already, that would increase then even more. Say thanks to some japanese jerks who linked to my FTPd here and don't take those links down... even if public access is refused for some weeks now...)

Humm. I hope these are the answers you were expecting :)

Best regards,
Koepi

Marc FD
9th September 2002, 18:06
Originally posted by Koepi

Your info about EPZS(^2) is misleading.
It's not better in any case. It's not worse in any case.

Or do you absolutly want EPZS^2 (who are not _better_) ?

when i said not _better_,
i wanted to point out you can't say wich is better.
but you're right, it wasn't clear.

David caid
10th September 2002, 08:07
Hello,

Yesterday, I installed the Umaniacs builds (XviD.Alpha.12.08.2002.1000.exe). It works perfectly. Virtualdub doesn't crash anymore (and encode perfectly).
So it is due to the compilation.

I can understand the bandwith (and the amount of work) trouble you can have Koepi. I just hope Umaniacs won't change his compilation mode.

Anyway, thank you all for solving my problem.

Ciao

David