[FFmpeg-soc] [RFC] Possible jpeg2k decoder rewrite

[Michael Niedermayer]

On Sun, Apr 26, 2009 at 05:47:11PM +0530, Jai Menon wrote:
> On 4/26/09, Michael Niedermayer <michaelni at gmx.at> wrote:
> > On Sun, Apr 26, 2009 at 04:17:48PM +0530, Jai Menon wrote:
> >  > Hi,
> >  >
> >  > So any opinions on $subject? Quite a lot of stuff is missing in the
> >  > current code (none of which is documented)
> >
> >  > and its also poorly
> >  > structured.
> >
> >
> > elaborate please
> 
> Well, there are just too many memcpys used to copy around data which
> should be maintaned globally. 

which memcpys exactly do you speak of?


> Also, supporting other progression
> orders requires a bunch of hacks because of bounds not being known.

elaborate


> 
> >  > Also I dunno how to add support for assembling out of
> >  > order tile parts etc to current code :) Suggestions welcome.
> >
> >
> > how many j2k files use out of order tiles?
> 
> Atleast p0_10.j2k in conformance suite does. My initial aim is to get
> atleast profile 0 working.
> 

> >  Why is it diffifult to support in the current code?
> 
> I'm not implying it is difficult. I just run into a few problems every
> now and then which is preventing me from coming up with a clean
> patchset :)

thats no argument for a rewrite


> That is why I thought of asking if a rewrite is a possible solution.
> Because after all, the code wasn't merged last time around so these
> problems existed back then too.

The last time around a project writing a j2k decoder (and encoder)
failed.
What makes you belive yours would succeed if you throw the existing
work away?
Of course in the end i want a j2k decoder and care little if its based
on the existing code or not.


> 
> Right now, the major problems (off the top of my head) are :
> 
> - only lrcp supported. should support atleast rlcp if not more.


> - POC marker is not read

the spec calls it POD or at least there is nothing called POC
and its not read because the other orders arent supported yet i suspect


> - component subsampling is not supported. this is required by a bunch
> of profile 0 test files. right now only 1x1 supported.

this must be fixed, but i dont see how thats a argument in favor of a
rewrite, and that applies to the other points here too

So as summary my oppinion is pretty much that iam against a rewrite unless
you can point at issues with the current code where a rewrite is less work
than fixing the issues in the current code.

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

While the State exists there can be no freedom; when there is freedom there
will be no State. -- Vladimir Lenin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-soc/attachments/20090426/9c1184f1/attachment.pgp>