Forums

Matlab fft and Intel MKL

Started by tadhgm August 4, 2007
Hello Everyone,

I am new to the site so bare with me. I am working on a college project
involving convolution operations on complex 2D data. The original code for
this program was written in Matlab and the C/C++ code that I am developing
has to match the Matlab Implementation. So far I have been able to
replicate the same data in Matlab except for the output from the Matlab
FFT. When I use the Intel 2D DFT and compare it's output to the Matlab 2D
FFT it does not match, yet it seems to match for the 1D case. Does anyone
know why this is and does the fact that I am using the MKL_Complex16 type
have anything to do with it.

Thanks,
Tadhg  


"tadhgm" <tadhg.j.mulkern@nuim.ie> wrote in message 
news:scadnaHxydWOKynbnZ2dnUVZ_tOtnZ2d@giganews.com...
> Hello Everyone, > > I am new to the site so bare with me.
Sorry! We don't know each other well enough ;-) (The idiom is, "bear with me.")
tadhgm wrote:
> Hello Everyone, > > I am new to the site so bare with me.
I do that only with intimate friends.
> I am working on a college project > involving convolution operations on complex 2D data. The original code for > this program was written in Matlab and the C/C++ code that I am developing > has to match the Matlab Implementation. So far I have been able to > replicate the same data in Matlab except for the output from the Matlab > FFT. When I use the Intel 2D DFT and compare it's output to the Matlab 2D > FFT it does not match, yet it seems to match for the 1D case. Does anyone > know why this is and does the fact that I am using the MKL_Complex16 type > have anything to do with it.
Compare the array layouts for the two cases. There is no standard, so you will probably need to do some transposing to see that the data match. The way they're scaled may also be different. Jerry -- Engineering is the art of making what you want from things you can get. &macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;&macr;