[vlc-devel] Re: Question about HD play

[chshin]

Hi. My name is c.h.shin, intern researcher of KJIST, south korea

>I have some questions:
thanks (-:

>> In documentation (VideoLAN client API) 4.6, it say VLC provides just an
MP at ML decoder for MPEG-2. Is this description valid until now?
>4.6? perhaps you mean 0.4.6? In that case vlc has droped it's internal
mpeg2
>decoder in favor of using libmpeg2. This was because libmpeg2 turned out to
>be considerably faster, and also better in many other ways.
Sorry, my expression was poor. I did not mean VLC version (My VLC version is
0.7.0), but chapter 4.6 of a doc ( VideoLAN client API).
The chapter say that VLC provide just a MP at ML decoder for MPEG-2. I'm
wondering VLC supports MP at HL specification.

>>
>> I already tested HD streaming that has 20Mbps bandwidth. Spec of test
systems run as follows;
>> maching for VLC
>> processor: ZEON 2.0 ( dual )
XEON - my mistake
>> memory: 1G byte
>> video: ATI FireGL E1 64Mbyte (dual)
>> os: winxp
>> display resolution: 2560 * 1024 (two monitors)
>Are you running two windows (using wall-filter or some other filter combo)
>or do you have one window stretching both monitors?
I have one window stretching both monitors. ( for almost full resolution
test )

>>
>> Test succeeded somehow. - good decoding, just few lost pictures but too
much occupation of  system resource.
>What system resources? CPU? memory? io? My laptop has a P4M cpu at 2ghz,
and
>it is far from able to decode such a stream. But my 1.8ghz athlon XP can
>decode it fine. I talked with the libmpeg2 author and he thought perhaps it
>has something todo with L1 cache. Do the XEON have more L1 cache than a
P4M?
I meant CPU as system resource. During stream is displaying, cpu usage rate
was over 50%.
My XEON has 12KB L1 cache. (maybe intel P4M has 8KB)

>> However, quality of display is not fine. It is main problem.
>I guess this is because the stream is interlaced and you didn't enable the
>deinterlace filter. This is available from the video menu in the windows
>gui. I suggest the "blend" mode for best speed/qualty tradeoff. Note that
>this will add a considerable extra cpu usage.
Great! I'm ashamed of my stupidity! Everything is being already supported by
VLC.
By set deinterlace filter "blend" mode, interferences disappear. Thank you

But some problems still remains a mystery.
For reference, I linked two pictures captured.

http://203.237.53.46/HDwihtVLC_patial_capture1.jpg
http://203.237.53.46/HDwihtVLC_patial_capture2.jpg

These pictures captured during displaying HD(1080i) source with VLC.
As shown in pictures, their quality is not good. There are many broken
pixels.
I guess the rectangles are macro blocks and  UV processing has some
problems.
On low display resolution ( not media resolution ), decoding, displaying
were done well.
But on high display resolutions (eg. 2560 * 1024, 2304 * 864 .. etc),
quality is poor as pictures linked.
Even decoding work was unstable.

>>
>> Is it possible to optimize VLC as my objective? So,  Can I acquire fine
quality screen?
>I doubt there is much to do. I've implemented some sse2-acellerations for
>the deinterlace, perhaps they will help (downlaod latest nightly build from
>download.videolan.org/~videolan/build to test). Except this, I don't see
any
>easy optimizations.

>Sigmund Augdal

-- 
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html
If you are in trouble, please contact <postmaster at videolan.org>