Fast FFT library function

[boazk]

Good morning !

1. The input dataof my application isa 32X16 complex floating point matrix.
2. The most demandingalgorithm (in terms ofCPU load and computation power) isthe FFT function, which is carried out on each row of the matrix (16 timesFFT on 32 complex floating point elements vector).
3. My factory purchased the IPP library, but I didn't find any library function which carries out an FFT on a matrix (not 2D FFT which is on of the library functions of the IPPI !!!) and the current FFT library function (included in IPPS) has a lot of calling overhead!
4. Please help me finding such a function (if available) or just give mea link to such a library which includes an appropriate one.

Thanks, Boaz

[Vladimir_Dudnik]

It seems youcreate FFTSpec for every FFT call.You should use only one Spec and one buffer for all 16 calls the calling overhead should not be visible via FFT processing.

Regards,
Vladimir

[michaeldunemedicalco]

he first you can use fft of image processing ippi

it called ippiFFTFwd

beware of the formats there are convertions.

also look at intel mkl library they have difrent ffts.

also i recomend you look in web FFT2D there are some algorithms. you can realy use fft of vector and to create of image.