IPP v4 speed ippcopy..ippMirror

IPP v4 speed ippcopy..ippMirror

In order to provide pixel swizzeling for a 2 tap camera i've used IPP with ippcopy() for first part of image ( halfline ) and ippMirror for second part of the image (halfline). This works well but is not as fast as expected. For a 24bit image 2048 * 2048 a trivial double loop in straight C is quite faster and less cpu time consuming. Why ? I have an intel board with 3Ghz P4 HT and 1Go Ram.
Anyway I really enjoy IPP and wait for upcoming book on it.
Sincerly yours
SRSR (france)

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

Hi, it looks a bit strange. Could you please provide us with your test code, so we will able to verify what is the problem. It is something that definitely should not be.



here is the function using IPP:

void ResequenceTap2_IPP_C3HLR(BYTE* pDest,BYTE* pSrc,long iWidth, long iHeight,long iPitchDest, long iPitchSrc)


IppiSize dstRoiSize;

dstRoiSize.width = iWidth /2;

dstRoiSize.height= iHeight;

ippiCopy_8u_C3R((const Ipp8u*) pSrc,iPitchSrc,


pSrc += iPitchSrc/2;

pDest+= iPitchDest/2;

ippiMirror_8u_C3R((unsigned char*)pSrc,iPitchSrc,

(unsigned char*)pDest,iPitchDest,dstRoiSize,ippAxsVertical);


Sincerly yours,

Thank you, we will try to reproduce it.


Book? What book?


Intel Press has to publish a book on IPP ( and an another on Hyperthreading ): take a look at Intel Press and, like me expect that the book announced for February 2004 will be soon here... perhaps in April ?

Leave a Comment

Please sign in to add a comment. Not a member? Join today