Intel® Integrated Performance Primitives
Deliberate problems developing high-performance vision, signal, security, and storage applications.
6704 Discussions

Ipp 9.0 ippiMul method doens't support 32fc_C1R

Daan_Zhu
Beginner
408 Views

I would like to update IPP from 7.0 to 9.0. But I found a method "ippiMul_" method (used to multiply two images) doesn't support complex value. I may find the support mode "32f_C1R". But I need "32fc_C1R" which is existed in IPP 7.0. Unfortunately I didn't find any replaced method in "Appendix C: Removed Functions for Image and Video Processing". It stopped me to upgrade IPP.

Could any one please advise me any replaced method or any walking around method?

Thank you very much.

Regards,

Dan

0 Kudos
3 Replies
Chao_Y_Intel
Moderator
408 Views

Dan, 

thanks for you report.  We have further check on this.  Meanwhile, could you check if IPP legacy package works for you? Here is some content on using and downloading this package:  
https://software.intel.com/en-us/articles/intel-ipp-legacy-libraries

let us know if it can work for you.

Thanks,
Chao

 

 

 

0 Kudos
Daan_Zhu
Beginner
408 Views

Hey Chao,

Thank you for your reply. "ippiMul" is already existed in ipp9.0 update 1 and 2. But just float complex type "32fc_" is not supported. Are there any reasons behind that?

Thank you very much.

Regards,

Dan

0 Kudos
Igor_A_Intel
Employee
408 Views

Hi Dan,

the simplest way to do this with IPP 9.x is to call ippsMul_32fc in a loop by rows - performance will be significantly better than with deprecated ippiMul_32fc from the previous IPP versions or legacy lib as ippiMul_32fc has not ever had any cpu-specific optimization, while ipps one is highly optimized for all Intel architectures.

regards, Igor

0 Kudos
Reply