I am using IPP6.0 in conjunction with opencv on a Penryn Core2Duo, and not seeing a significant speedup. I see that the appropriate ipp functions are being called, but the performance remains about the same. Could this be because the data is not in the cache at the time? How can I ensure that the ipp functions will have an effect on my runtime?
I build ippopencv (which comes with opencv source) and am using that in conjunction with OpenCV. I see that ipp is recognized and that the cpu type is recognized as well. I am comparing OpenCV calls, specifically looking at cvErode and cvDilate with and without IPP. I don't see any performance differences between the 2 runs. Any suggestions?