topic Re: Complex numbers with Intel IPP in Intel® Integrated Performance Primitives

Complex numbers with Intel IPP

haintz — Tue, 06 Nov 2007 23:30:30 GMT

It looks like only a few funtions in the Intel IPP can deal with complex numbers. For example it looks like SVD only deals with real matries and not complex matrices.

Is it better to use MKL for complex matrices rather than IPP or am I missing something in IPP?

Thanks,
Pavel

Re: Complex numbers with Intel IPP

Vladimir_Dudnik — Mon, 12 Nov 2007 16:23:01 GMT

Hi Pavel,

many functions in IPP signal and image processing domains are work with complex numbers.

Note, IPP small matrix domain was designed to cover operations with small matrixes. The typical sizes are in 3x3, 5x5, 7x7, 11x11 range. If you need to process 1000x1000 matrixes MKL will be the best choice.

Regards,
Vladimir

Re: Complex numbers with Intel IPP

its4soo — Wed, 07 May 2008 11:15:33 GMT

HI ,
How can I use Cholesky method for Complex Matrices with IPP? If not possible can you suggest some alternate method?

Regards
Soo

Re: Complex numbers with Intel IPP

Vladimir_Dudnik — Thu, 08 May 2008 22:59:23 GMT

Soo,

if you look for matrix operations on complex data types then we suggest you to consider MKL as a solution. In IPP, small matrix operations do not support complex data types (at least in current version and there is no short term plans to add such functionality)

Regards,
Vladimir

Re: Complex numbers with Intel IPP

its4soo — Thu, 15 May 2008 08:00:43 GMT

HI,
Thanks for your reply... But where can find some good examples to start programming C++ with MKL fro complex matrices? Whether the MKL is also sacalable as IPP among different intel processors?
Regards
Soo