JpegView Performance

[hyang]

I test the jpegview on both P4 2.4GHz (without HyperThread) and Dual Xeon 3.2GHz (with HyperThread) PC and found out jpegviewrunningaboutfour time fasteron encoding1024x768JPEG file. The compiler I am using is VC++ 6.0. Why? Do I have to use Intel C/C++ compiler? Thanks!

[Vladimir_Dudnik]

Hi, so what the question? You find that codec runs faster, that is good, is not it? You are right, if you will compile it with Intel compiler, it can be even faster. Note, JPEG decoder can be threaded with using OpenMP (just compile it with Intel compiler) but encoder does not contain threading yet.

Regards,
Vladimir

[hyang]

Sorry I didn't put my question clearly. According to my test P4 2.4GHz is faster thendual Xeon 3.2 GHz on encoding 1024x768 JPEG file. I also disable Hyper Thread on dual Xeon PC and notice that JpegView running faster but still slower than P4 2.4 PC.

My question is do I have to use Intel c/c++ compiler and other library in order to take the advantage of IPP. Is there anyone out there have the similar problem? I am using VC++ 6.0 and IPP 5.0. Thanks for help!

[Vladimir_Dudnik]

Hi,

Please note that JPEG encoder part in JPEGView sample was not threaded yet, so you shoul not expect speedup from dual-core or SMP systems from it. You can look on decoder part which is threaded with using OpenMP, it should show clear benefits from using multiple-cpu systems. Note, effectiveness of threading in current implementation is different for different JPEG sample factors and is most visible (about 1.7x) for 422 sampling factors (mostly used in Motion JPEG streams. We are plan to add threading to encoder part as well, but I can't refer you to any particular time when it will be released.

Regards,
Vladimir