IPPISet is slower than memset & IPPSSET

IPPISet is slower than memset & IPPSSET

murali_523's picture

Hi,

  We tried IPPISET roi based method to set the set of portions in the image 0.

  But this is slower than memset [on entire image to 0]

  We have tried the IPPSSET and found it is more faster than memset and IPPISET.

  Is the ROI based method usually slower than non-ROI based method like IPPSSET.

  Actually we thought since we were doing selective regions in image to reset to zero using IPPISET ,it should be faster than memset and IPPSET   performed on entire image.

   But when we profile we found the IPPISET[called four time on four different small region on image] is taking more time than IPPSSET and memset.

  Can you please explain why/how this can happen?

  We were using IPP 7.0 version.

 Thanks & Regards,

Murali

6 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.
Sergey Kostrov's picture

Some time in 2012 I've evaluated performance of ippsSet_8u and ippiSet_8u_C1R on an image with size 34207 x 34207. However, it was for the entire image without using ROI(s).

>>...Is the ROI based method usually slower than non-ROI based method like IPPSSET...

I wouldn't make any statements until some tests are completed. Could you post your test-case? What ippsSet_xxx function did you use?

>>...it should be faster than memset and IPPSET performed on entire image...

Please see my comment above.

Sergey Kostrov's picture

Update:

>>...What ippsSet_xxx function did you use?

What ippiSet_xxx and ippsSet_xxx functions did you use?

murali_523's picture

For IPPsSet i had used ippsSet_8u

For IPPISet i had used ippiSet_8u_AC4

The image is a 32-bit color image and we need to set the four corner ROI's to '0' value

Sergey Kostrov's picture

>>...The image is a 32-bit color image and we need to set the four corner ROI's to '0' value...

Could you provide more detailed details, I would say so?

Igor Astakhov (Intel)'s picture

AC4 should be slower by definition - AC4 memory layer means 4 channels with A channel - A channel in IPP is not touched by definition. To set all 4 channels to 0 you should use ippiSet_8u_C4R.

Regards, Igor

Login to leave a comment.