[Vorbis-dev] Complaince testing for oggvorbis encoder

[Michael Smith]

On 2/24/06, parul <parul.agarwal at einfochips.com> wrote:
> We are working on OggVorbis encoder. In the porting effort we are trying
> to convert it to fixed point code (both 32 and 24 bit fixed point). Now
> the issue is how we do the testing. What should be the criteria for our
> testing. Does anybody has any idea how compliance testing (i.e. some
> objective tests) is done at encoder side? What is the criteria of
> testing at the encoder side? If anybody could help?

As with most other codec specifications (such as the mpeg ones),
vorbis doesn't define encoder compliance - the only compliance
criteria are that it produces valid, decodable, bitstreams. Of course,
it's desirable to also produce something that sounds close to the
input, but that's not, strictly speaking, required.

For testing that your bitstreams are valid, you can use the 'vorbose'
program from xiph SVN, which is designed for this sort of thing - it
doesn't acually do decoding, it just validates a bitstream.

Testing with existing decoders is also a useful, simple test.

Mike