[MPlayer-users] Re: [POOL] Moderate some users (was: Re: [-] Re: divx 6)

Matthias Wieser mwieser at gmx.de
Fri May 5 10:22:31 CEST 2006


Am Freitag, 5. Mai 2006 02:21 schrieb RC:
> On Thu, 4 May 2006 21:55:23 +0200
>
> Matthias Wieser <mwieser at gmx.de> wrote:
> > Please elaborate a bit more.
>
> Well, at the risk of possibly starting another flamefest...
>
> For starters, I (and someone else IIRC) mentioned early in the thread
> that a big part of the performance difference was because lavc was used
> single-threaded, while Divx and Xvid were not, yet you continued to
> ignore that,

That's completely wrong, read the doom9 comparison:

"XviD 1.1 is finally about to be released. A few bugs have been fixed since 
beta2, and recently there has been some SMP optimization, but the release 
of those optimizations was too close to the deadline to make the test."
(http://www.doom9.org/index.html?/codecs-main-105-1.htm)

They tested lavc singlethreaded because that have been the options suggested 
by the mencoder and lavc developers. I remember that especially one person 
was completely against multithreaded lavc encoding because it leads to 
marginally worse quality (probably invisible).

> and repeat that lavc is: "several times slower than xvid or 
> divx" when the performance difference, at the very least, is really much
> closer.

No, it's not. They tested lavc single threaded against xvid single threaded. 
Xvid has been several times faster (71.43 fps vs 26.32fps see URL above) 
while giving better quality. Using xvid with multi cpu support would have 
made it even faster. And using higher quality lavc options would have made 
lavc even slower. That are just facts.


> When talking about GMC, you just say "It can improve quality" without
> getting into any specifics to make your point.

Of course it can help. Not only scientific animated visualisations but also 
"real world" content. E.g. I have encoded a video captured from a 
helicopter with and without gmc.
gmc gives 35.50dB vs 35.44 dB without gmc. This clearly proves that gmc can 
help (1).

> No mention of what type 
> of content it helps, how much it can help, etc.  Not conducive to a
> rational discussion at all...  just a shouting match.

?? I don't want to post a gmc tutorial. I just mentioned gmc as one of 
xvid's features. If someone want's to start a discussion about the 
uselessness of gmc *he* should post numbers, not I.


>
> When Rich mentions some quality issues discussed in the doom9 thread,
> your only response is "Haha! Which ones?"

Exactly. Up to now he has not posted one single url ore something 
comparable. Yes there have been minor quality issues but they had no 
significant influence on the result. If he does not believe this he should 
post urls which show the contrary.

> Then you make some vague challenge.

The Doom 9 testing procedure is fully transparent. If he thinks that they 
made some mistakes he can do an own encoding (e.g. with mencoder) and show 
that this would have changed the result.
The used lavc options have been (see doom9 homepage):
 vcodec=mpeg4:vbitrate=560:mbd=2:mv0:trell:cbp:precmp=2:cmp=2:subcmp=2:\
  predia=2:dia=2:preme=2:v4mv:last_pred=2:vqcomp=0.6:vmax_b_frames=2:qpel:\
  vpass=1:turbo
the xvid options have been: MPEG quantization, QPel, VHQ for B-frames
xvid version: XviD 1.1 CVS build dated December 19th
Libavcodec: CVS build dated 11/30/2005

So there is nothing vague in there.

> > These have been three individual responses. Please quote fully and
> > correctly.
>
> Yes they were.  Anyone who wants to read your full email can scroll-up
> or visit gmane, I'm not going to quote the whole thread.
>
> > Calmed down? :-) I have asked several times to post to my privat email
> > address. BUt never got one of those flame mails. :-(
>
> No, and you don't seem to have sent your e-mails privately to him,
> either.

Of course not, because I don't want newbies reading this mailing list (or 
the archives) believe that xvid and divx are slow and bad and that the 
reason for the not-so-good lavc results in the doom9 contest have been 
caused by unfair rules.

Regards,
  Matthias

(1)
xvid options: vhq=4:bvhq=1:chroma_opt:quant_type=mpeg:bitrate=400:[gmc]:psnr

without gmc:

Video stream:  397.936 kbit/s  (49742 B/s)  size: 7963368 bytes  160.093 
secs  4802 frames
Audio stream:  136.020 kbit/s  (17002 B/s)  size: 2720402 bytes  160.000 
secs
The value 99.99dB is a special value and represents the upper range limit
xvid:     Min PSNR y : 28.29 dB, u : 66.10 dB, v : 66.10 dB, in frame 3827
xvid: Average PSNR y : 33.44 dB, u : 60.91 dB, v : 60.91 dB, for 4796 frames
xvid:     Max PSNR y : 45.35 dB, u : 99.99 dB, v : 99.99 dB, in frame 63

with gmc:

Video stream:  398.081 kbit/s  (49760 B/s)  size: 7966265 bytes  160.093 
secs  4802 frames
Audio stream:  136.020 kbit/s  (17002 B/s)  size: 2720402 bytes  160.000 
secs
The value 99.99dB is a special value and represents the upper range limit
xvid:     Min PSNR y : 28.35 dB, u : 59.59 dB, v : 59.59 dB, in frame 3828
xvid: Average PSNR y : 33.50 dB, u : 61.11 dB, v : 61.11 dB, for 4796 frames
xvid:     Max PSNR y : 45.35 dB, u : 99.99 dB, v : 99.99 dB, in frame 63




More information about the MPlayer-users mailing list