Intel® Integrated Performance Primitives
Community support and discussions relating to developing high-performance vision, signal, security, and storage applications.
Announcements
This community is designed for sharing of public information. Please do not share Intel or third-party confidential information here.
6628 Discussions

Wrong result when using CrossCorrValid_Norm

panda2420
Beginner
123 Views

Hello.

We have received wrong results while using CrossCorrValid_Norm with IPP version 5.3 update 3.

The results are not always wrong, but for certain matrices (with certain size) , the results diviate for some points by 0.3. It doesn't matter what is the data in the matrix- Just the size. We have used previous versions of IPP and the results were always correct.

in ourtestings, this happens only while using UINT8 matrices(Images).

For Example: For two matrices: one sized 3X5 and a smaller one 2X3.

3X5:

85 194 177 253 39

241 233 2 32 19

23 250 93 243 132

2X3

149 84 17

2 227 19

IPP Result

0.121069 -0.616973 0.0344648

0.539306 -0.28021 0.162498

Ours and Matlab results

0.19392 -0.513159 0.0344648

0.578099 -0.178333 0.162498

As was said before , It does not matter what the data in the matrices is, the result always diviates.

Thank you

Dana O.

0 Kudos
1 Reply
Vladimir_Dudnik
Employee
123 Views

Hello Dana,

Thanks for reporting this. We identified a bug in optimized version of this function. The bug will be fixed in the IPP 6.0 final release. Note, the latest availbale version,IPP 5.3 update 4,still contain that issue.

Regards,
Vladimir

Reply