OpenCV+IPP 2.: Profiling hell

I would like to profile what is going on with this, because it would be very nice to use this, because it is uses both CPU core parallel, thus I should not take care about run two different process to take advantage from multiple cores in the future.
I spent some hours to profile my library, because sprof said after run FindShapes with profiling the library:

sprof: failed to load shared object `libmyLibrary.so.0': No such file or directory

I tried everything and it was very suspicious that google said nothing about it.... After reading some additional articles about shared library profiling I found that the libraries and the program being run should not be compiled with profiling feature, because it confuses the profiler:

"...For the same reason, these applications must not be compiled with the -xpg option of cc..."

Why do the man pages related to profiling not say anything about it?...Mysterious... Search futher...Found: gprof does not support multi-threaded applications and "gprof is useless in today" from Ulrich Drepper, its author in 2002. Hey!? What the fuck'n shit? Out-of-date and it man page misses important points of its usage, then throw it in the trash. Please.
Thus, after recompiling everything without profiling, trying to use gprof and sprof without success, the last chance: oprofile and I have luck, there is a nice UI made by OpenedHand and repository for feisty/gutsy. I installed and it works, great! :)
Thus what is the profiling result?...

Optical flow gives random result...

I tried two kind of optical flow algorithm in OpenCV: cvCalcOpticalFlowHS and cvCalcOpticalFlowLK, but both give me random velocity matrices...something is bad....

OpenCV+IPP, what the heck?

I tried yesterday evening to use IPP 5.3 with the latest cvs version of the OpenCV, but the result was ridiculous: the computation time remained the same, but my program used both two CPU cores....funny.

Kupac

Ime, ilyen egy hétvégi kupac, ha az apuka nemtetszőt szól be a csajoknak... :)

Fixed V4L frame rates in OpenCV

It was really ridiculous. After 30 minutes dig into the sources, I found that the OpenCV forced the v4l device to switch to NTSC mode. That was the reason of the half frame rate in v4l, let's commit it...

First task is done. :)

Thinking on next steps

Since I finished my work on background subtraction, the question comes up what to do next. There are some start point to continue:

- Check the v4l/v4l2 capture codes of the OpenCV, because it is really ridiculous, that the OpenCV can capture only with half frame rate. Need to fix it.
- Read some papers from biologic point of view about feature point search on images to track objects.
- Embed the optical flow techniques of OpenCV to myLibrary and integrate with background subtraction to handle the self-movements of the camera.
- Update my project home page on aibo.sf.net and probably create a new release....

My new paper

Finally, I sent my first publication to a "serious" conference about foreground subtraction. I'm very proud, because that is an important milestone in my research.