Intel OpenCV Evaluation Guidelines?

Question

We have quite a lot of openCV code, and we plan to investigate Intel IPP and see what accelerations we can achieve in our application.

Intel has a white article on how to bundle OpenCV and IPP. Using Intel IPP with OpenCV It seems that the link may be automatic in some cases, but the scope of support is unclear.

We are a Windows store and I am interested in any war stories or problems / issues that should be considered when conducting such an assessment.

thanks

Answer 1

More and more less OpenCV has something to do with IPP. In subsequent years, IPP can hardly provide anything different from the results of GCC or any other worthy compiler on Intel platforms. In the old days, IPP provided up to 50% acceleration, but not more if we look at all the features of GCC, such as TBB.

I have been reading OpenCV lists for many years, and I cannot recall that anyone has talked about IPP for a long time. A search in my OpenCV mailbox also showed nothing. I suspect that IPP support is already not so tested and can even be very broken.

I saw huge improvements when manually setting up OpenCV on platforms other than Intel. For example, on iOS, I got a big boost, replacing cos () and sin () with lookup tables. But IPP is for Intel only.

The new trend seems to be GPU processing, and there you can see the likelihood of being 10 or 20 times faster. Engineer time is also expensive, and instead of looking for hard work to get 20% profit, invest in a GPU if applicable.