Re: [video-codec] Performances, measurement of same (Re: The Can Has Landed)

[Harald Alvestrand <[e-mail]>]

On 03/23/2015 01:52 PM, Mohammed Raad wrote:
> Well, the second part (how one should judge the potential
> optimization) typically has been avoided in video standardization and
> so far there seems to be little willingness to work on the development
> of a metric or a set of metrics that can be used to determine how
> "optimizable" a codec or tool is. It would actually be good if people
> at the IETF can develop such a metric.

I was thinking more of whether experience with the encoding speed of the
test implementations should be allowed as an argument or not.

Sometimes one can use dataflow analysis and similar tools to show that
certain techniques prevent optimization beyond a certain level (or not)
- but most of the time, I think "this runs at speed x / quality y" is a
more compelling argument than "I think this could be optimized to speed
x / quality y".

>
> Also, keep in mind that the incremental development of a codec means
> that prior decisions limit what can be done with tools proposed later
> in the process.

Yes. Especially when the decision is "we can't use this tool for
nontechnical reasons", it means that we can't accept anything that is a
refinement of this tool either.