theora dev - Oct 2002 - Simple player and encoder binaries for Windows

--- Stephane wrote:
VP3 codec uses a lot of CPU for decoding. We will need some optimization
because now this codec is way too slow for home uses.
--- end of quote ---

Actually, the Windows VP3 QuickTime codec runs just fine on a 400MHz Pentium II
box, with only occasional glitches on some dodgy Celeron systems.  At work, we
routinely test 320x240 full-color, full-motion video on low-end boxes.  At
300MHz, VP3 displays some glitches; at higher speeds, most systems are generally
fine.  (The VP3 QuickTime codec uses the native assembly code supplied with VP3,
not the much slower C code.)

Please do not assume anything about the performance of VP3 from the performance
of Mauricio's sample code.  Such assumptions would be wildly inaccurate, for
reasons Mauricio has already explained.

Cheers,
Eric
--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to
'theora-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.

note that there may be some changes to the Theora algorithm, ie it is a new
release, that's why it's not called VP3.
 
The plan was to get the bitstream down, then start on optimization.  I'm
sure the original VP3 asm libs will be of great value but they will probably not
work as-is.  Until optimization time, please do not assume anything about
performance.

        -----Original Message----- 
        From: Mauricio Piacentini [mailto:mauricio@tabuleiro.com] 
        Sent: Tue 10/29/2002 12:10 PM 
        To: theora-dev@xiph.org 
        Cc: 
        Subject: Re: [theora-dev] Simple player and encoder binaries for Windows
        
        

        Just fyi there are already assembly versions of the critical functions
        available in the vp3 codec tree, for Windows and Macintosh. The current
        theora code however uses the slower C functions, for portability. I
believe
        in the future (after 1.0 or very close to it) it may be possible to add
        these assembler files back to the tree at least for platform-specific
        builds, but this is the kind of work that really belongs to the end of
the
        cycle.
        I don't think VP3 is a slow codec. It is too early in the theora
cycle to
        tell what the real performance will be, and where to optimize. If the
VP3
        code base is any indication it can perform very fast even on relatively
low
        end systems (PII 300 or so.)
        
        Regards,
        Mauricio Piacentini
        Tabuleiro
        
	> It could be recoded in assembler, but I hate to
	> do that if there are major changes to be made...
        
        
        --- >8 ----
        List archives:  http://www.xiph.org/archives/
        Ogg project homepage: http://www.xiph.org/ogg/
        To unsubscribe from this list, send a message to
'theora-dev-request@xiph.org'
        containing only the word 'unsubscribe' in the body.  No subject
is needed.
        Unsubscribe messages sent to the list will be ignored/filtered.
        

-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 5254 bytes
Desc: winmail.dat
Url :
http://lists.xiph.org/pipermail/theora-dev/attachments/20021029/5c67214c/winmail-0001.bin

on my machine (800 Mhz celeron laptop), I have two problems with playback using
this binary:

1) image flickers (some frames are black).  To be fair, I've seen this
problem with other players, ie RealPlayer and On2 players, so I wouldn't
worry about it, it's probably a bug in my driver
 
2) sound cuts out at around 2 sec into piece.  From my days programming WaveOut,
that sounds like you're getting one pass through your (circular?) buffer. 
Or, if you're chaining buffers, somehow it isn't getting initialized
correctly.
 
Looking forward to the code - dan

        -----Original Message----- 
        From: Mauricio Piacentini [mailto:mauricio@tabuleiro.com] 
        Sent: Mon 10/28/2002 3:09 PM 
        To: theora-dev@xiph.org 
        Cc: 
        Subject: [theora-dev] Simple player and encoder binaries for Windows
        
        

        Binary, readme and test files at
        
        http://www.tabuleiro.com/mauricio/theorawin.zip
        
        The player uses some video routines from player_example.c, but with a
        different audio and synch mechanism, built on top of PortAudio. This one
was
        built with support for the Wave out device, but it is easy to change it
to
        use DirectSound (or OSS in Linux and SoundManager/CoreAudio for the
Mac.) I
        am planning to port this to Linux/Mac as well shortly. I will post the
        sources when it is done, or let me know if you want them in the current
        state (currently VC++/Windows specific, but I plan to clean them up and
add
        at least a Mac project.) They also depend on the small patches I sent
last
        week to the list, so I may need to post a complete working tree instead
        while the patches do not show up in CVS.
        
        I compiled the encoder as well, a quick port to remove some Gnu/linux
        specific functions not present on Windows (like rint() and the getopt
        package) and add some headers.
        
        Anyway, I hope these may be useful for a quick test until I have time to
        clean up the sources. Let me know if you find problems.
        
        Regards,
        Mauricio Piacentini
        Tabuleiro
        
        --- >8 ----
        List archives:  http://www.xiph.org/archives/
        Ogg project homepage: http://www.xiph.org/ogg/
        To unsubscribe from this list, send a message to
'theora-dev-request@xiph.org'
        containing only the word 'unsubscribe' in the body.  No subject
is needed.
        Unsubscribe messages sent to the list will be ignored/filtered.
        

-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 5818 bytes
Desc: winmail.dat
Url :
http://lists.xiph.org/pipermail/theora-dev/attachments/20021029/e740d4c9/winmail-0001.bin

Yes, as I mentioned before there is no trace of optimization going on at
this time, since theora is still alpha. It is unfair to try to compare
performance with other codecs at this stage.

This said :) I ran some simple tests on my machine, which is a fast Athlon
XP 1900+. Even at the current stage the simple player can decode audio and
full DVD-quality video (720X480) at very acceptable rates. I can decode 125
seconds of  this high resolution source material in about 38 seconds on this
setup. To compare, the sample code that blits the pixels to the SDL surface
takes more cpu time than the whole decoding of video and audio. Even with it
enabled I can still decode the same 125 seconds source in about 80 seconds
or so, so there is approx 30% cpu left.

Regards,
Mauricio Piacentini
Tabuleiro
> Please do not assume anything about the performance of VP3 from the
performance of Mauricio's sample code.  Such assumptions would be wildly
inaccurate, for reasons Mauricio has already explained.

<p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to
'theora-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.