[Ffmpeg-devel] Optimised yuvj420p_to_rgba32

[Michel Bardiaux]

I succeeded in decoding from my application a simple JPEG file using 
lavc (details on request). Its 40% faster than using the generic libjpeg 
(IJPEG6b), which is great. But -pg profiles show that 40% of the time is 
spent not in decoding the image, but in converting to B/G/R/X which is 
mandated by the rest of my application since the image has to be sent to 
an X server on Ix86/Linux.

Is there a hope of SSE-fying that function? I mean, I dont ask that 
anyone do it, just that someone expert in MMX and SSE tells me if it is 
*doable* (and worth it).

Greetings,
-- 
Michel Bardiaux
R&D Director
T +32 [0] 2 790 29 41
F +32 [0] 2 790 29 02
E mailto:mbardiaux at mediaxim.be

Mediaxim NV/SA
Vorstlaan 191 Boulevard du Souverain
Brussel 1160 Bruxelles
http://www.mediaxim.com/