CCS FORMAT FOR FFT

[Silva__Rafael]

Hi Guys, Im new here, so sorry if Im posting on the wrong place.

Im trying to use the FFT functions from MKL, and as far as I understood, the answer of the forward FFT is on CCS format. I need to draw a picture (because in my application this is an FK spectrum), so I want to have this matrix as a regular mxn matrix with the values, not in a different format like CCS.

Is there any option I can use or some helper function to do that? Or should should I do it by hand?

PS: I know is not so difficult, but this feature can help to speedup the software development.

[Silva__Rafael]

Ive implemented the code to unpack, and tested with a diagonal matrix 8x8 matrix, filled with "1" on the diagonal. The result should be(according to matlab):

8.0 0.0 0.0 0.0 0.0 0.0 0.0

0.0 0.0 0.0 0.0 0.0 0.0 8.0

0.0 0.0 0.0 0.0 0.0 8.0 0.0

0.0 0.0 0.0 0.0 8.0 0.0 0.0

0.0 0.0 0.0 8.0 0.0 0.0 0.0

0.0 0.0 8.0 0.0 0.0 0.0 0.0

0.0 8.0 0.0 0.0 0.0 0.0 0.0

0.0 0.0 0.0 0.0 0.0 0.0 0.0

But I didnt get it even unpacking the CCS. I got this:

8.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

0.0 0.0 0.0 0.0 0.0 0.0 8.0 0.0

0.0 0.0 8.0 0.0 8.0 0.0 0.0 0.0

0.0 0.0 0.0 0.0 8.0 0.0 0.0 0.0

Does anyone know if there is something missing?

[Dmitry_B_Intel]

Hi,

Irecommend using CCE format for 2-dimensional real-to-complex/complex-to-real FFTs. The data layout forit is much simpler to comprehend and code than 2D CCS. And, what we call CCEis also supported by other FFT libraries. Examples for this format can be found in MKL installation, directory examples/dftc and other. Recent update MKL 10.3.5 contains examples that deal explicitly withthe2D layouts, includingCCE, CCS and other formats.

Do not hesitate to post here in case of any kind ofissues with MKL.

Thanks
Dima