That looks like I might be able to get it to work, but it I'm not sure. I'm not understanding the documentation well.

However, you gave me an idea to wonder if I could do this with a shift. After downloading another help file, I found ippsLShiftC_32s_I function. If I run this shift 32bits left for 8 bits, I should get the pattern I need.

Thnaks for suggestion.

Hi,

Use ippiCopy_8u_AC4C3R, refer to ippiman.pdf, refer to Image Data Exchange and Initialization Functions

http://software.intel.com/sites/products/documentation/hpc/ipp/pdf/ippiman.pdf

Regards,

Naveen Gv

Hi Tom,

You have received the best answer from Naveen Gv (Intel).

For additional information, I have an extended discussion with Artem Vorobiev (Intel) please read this thread:(ippiYCbCr444ToYCbCr420_8u_C2P3R) is there any? + Y211)

URL: http://software.intel.com/en-us/forums/showthread.php?t=71620&o=d&s=lr

Regards,

Tamer

If I understood it right, the problem is that the alpha channel is on the first byte - but ippiCopy_8u_AC4C3R assumes the alpha channel on the fourth byte. The A0C4C3R channel format isn't supported in Copy function. Furthermore it can't be executed inplace. So I think shifting or swapping channels is the more correct way. And both can be done inplace if speed is needed. According to ps_ippi.exe SwapChannels seems to be a little bit faster than shifting.

Thanks for update, that is sort of what I was wondering. I think I'm going to go with shift/swap channels, then copy in final program.