[FFmpeg-devel] [PATCH 2/3] Indeo 5 decoder: common DSP functions

[Kostya]

On Sun, Jan 17, 2010 at 06:58:51PM +0100, Michael Niedermayer wrote:
> On Sun, Jan 17, 2010 at 04:58:23PM +0200, Kostya wrote:
[...]
> >  
> > It's not avoidable - this scaling cannot modify input since it's used
> > for further decoding and output is uint8_t, so it is simply not enough
> > for holding intermediate values.
> 
> static void horizontal_compose53i(IDWTELEM *b, int width){
>     IDWTELEM temp[width];
>             ^^^^^^^^^^^^^
> this is avoidable

and some people will definitely ask you to avoid it because
variable-length arrays are a bit controversial
 
> > In theory it should be faster, but unfortunately is not straight
> > applicable here.
> 
> If we cannot merge the 2 5/3 wavelets then then optimizations would also
> be duplicated (assuming there are volunteers for both) this feels like
> a bad thing to me

Don't worry about volunteers - up to now we have neither for both :(
IIRC, those two transforms also have inverted signs in lifting step too,
which compilcates things a bit too.

> But if its really not possible then its not a reason to hold up indeo

Not without unpredictably serious changes to both codecs.

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