[vorbis] Compression Artifacts at -q 5. Help!

[John Morton]

On Tue, 17 Jun 2003 11:52, Nemo -earth native- wrote:
> On Mon, Jun 16, 2003 at 04:13:33PM -0700, Christian.Buchner did utter:
> > Also the stream's serial number is a random number which would prevent
> > using the MD5 algorithm on the file.
>
> That's settable via commandline in oggenc, at least.
>
> Floating point rounding as mentioned before sounds like it would kill
> md5sum as a method - so you'd need to have an designed-for-this-task
> hash that basically understands ogg/vorbis files and can accomodate
> such things.
>
> I don't think anyone is likely to consider it a valuable use of their
> time though.

Just set the testing up to be done subjectively. Include a test pcm file (a 
.WAV or whatever), and the output .WAV created by a compress/decompress cycle 
on the test file with oggenc/dec built using a know-to-be-good compiler. Then 
have a 'test' target that does a compress/decompress on the test file with 
the newly built tools and throws you into an ABX tool to check for any 
obvious bad compiler generated artifacts. 

A very handy tool for binary package builders, especially as they can be 
asked to submit bad compiler bugs to the tracker, and those reports can be 
summarized into future install documentation.

John
--- >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 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.