IPP v4 G.729 Sample Question

IPP v4 G.729 Sample Question

Hi;
I just test the G.729 Encode/Decode sample and I found interested results:
The Computer is a Pentium 4 2.4 GH ( no HT ) running Windows 2000.

Whe I run the sample as is ( using the provided code and the DLLs ) I got for a PCM of 6 seconds:
Encode: ~25 Mh
Decode: ~ 7 Mh

But when I compile it with the "Single processor method" and linking with static libraries ( ipp*merged.lib ) using px, w7 or even t7 I got:
Encode: ~56 Mh
Decode: ~25 Mh

Why the DLL code is faster than the static-lib code ?

It is normal that using the static-libraries the G.729 Encode/Decode exe files are around 1MB ? ( The same using the DLL are around 50KB )

I'm evaluatng the IPP libraries to use them for voice compression for a VoIP client software that will run on Windows, Linux and PDAs, and the size of the code is very important. We are evaluating other libraries and the size of the code ( for example encode/decode 729 ) is aroung 90KB. We also will try it for Vide encoding ( H.263, H.263+ ), Where can I found the H.263 Encode Sample ???

Thanks

5 posts / 0 nouveau(x)
Dernière contribution
Reportez-vous à notre Notice d'optimisation pour plus d'informations sur les choix et l'optimisation des performances dans les produits logiciels Intel.

HI..I had the opportunity to work optimizing G.723.1 and G.729.1 to reduce proccessing and to reduce the size of the code..I worked in asm. Basically you must know the arquitecture of the processor o the types of chips where you desire to run your codec..You can reduce the size and to improve the yield.

H.263 you find in ITU..

best regards

Hi,

H.263 video encoder functionality is supported in MPEG4 encoder sample provided with IPP. Please take a look on it. You need to set some parametersin .PAR file to force MPEG4 encoder to generate H.263 bitstream.

Regards,
Vladimir

I would expect H.263 encoder to be much simpler and easier to read than MPEG-4 one ( especially if you have noISO standards ).Maybe someone extracted pure H.263 encoder from the IPP sample ?

I use a static libs'ippiemerged.lib', 'ippimerged.lib'. Do Ineed to worry about performance degradation when using static libs ?

Hi,

we do not provide separate H.263 encoder sample in IPP v4.0. If you find it useful you can submit this as a feature request through the Intel Premier Support service.

What about performance degradation when you were using static merged libraries - it seems for me that you did not call ippStaticInit/ippStaticFree functions, they are needed to initialize static dispatcher. Without it dispatcher will select generic PX code.

Regards,
Vladimir

Laisser un commentaire

Veuillez ouvrir une session pour ajouter un commentaire. Pas encore membre ? Rejoignez-nous dès aujourd’hui