[FFmpeg-devel] DCA Decoder

[Michael Niedermayer]

On Wed, Mar 11, 2015 at 09:44:22PM +0000, Derek Buitenhuis wrote:
> On 3/11/2015 9:36 PM, Michael Niedermayer wrote:
> > Thats analogous to saying "no its not important to put fuel in a
> > car, its important to drive the best car"
> 
> No what I propose is to look at both and decide which is best. Simply being
> submitted to FFmpeg first does not make it better.

you continue to talk about something completely unrelated to what i
said/meant.

you: take best code

I: for code to be ever in FFmpeg it must either be developed on top
of FFmpeg or it must be rebased/ merged/integrated at some point. Its
better if its developed and tested on top of FFmpeg in the first place
as this is less work and has a lower chance of bugs.

The difference here is the "question" that is awnsered
like in "there are 2 X implementations, which should we use" vs.
"I want to write a X implementation, on top of what codebase should
 i do the work"


> 
> > the 2nd is more work, so i suggest that new code is based on top of
> > FFmpeg already. Merge/rebase that patchset from Libav if you want
> > to work on top of it.
> 
> Yeah, merge one patchset in FFmpeg, and if it turns out the other functions
> better, the resulting mess is best summed up as "a clusterf*ck"...

you really are missunderstanding what i meant

I did not mean to suggest to push
something to main FFmpeg master, i suggested or intended to suggest,
that code be developeed on top of FFmpeg, code generally is developed
by a devloper locally on his box or in his own public repo.


> 
> > The more code is rebased and merged around the higher the risk of
> > bugs
> 
> This is a strawman argument (or perhaps just FUD).

From your point of view its indeed a strawman, and from my point of
view your arguments are a strawman because we just talk about 2
completely different things from the begin.


> 
> > Also if someone has testcases for all the new DCA features, i would
> > be interrested to have them so i can test these things if/when needed
> 
> From what I understand, neither of them implements fixed point yet in
> the core DCA decoder, which means neither is actually lossless or
> bit-exact.

ok, but do they implement something testable ?
and if so, how can this be tested ?
knowing this will be usefull to anyone working on the code

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

I do not agree with what you have to say, but I'll defend to the death your
right to say it. -- Voltaire
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20150312/7f03a4be/attachment.asc>