I'm implementing convolution using FFT and am looking for faster implementations.I had a guestion about algorithm used in FFT (real samples) for IPP . Is it possible to know what algorithm it uses (radix 2/4, split radix ) ?
I recalled, Intel IPP FFT is Radix - 2 FFT and the DFT have arbitrary radix support. Additionally, Intel Math Kernel Library (Intel MKL) has none Radix-2 DFT functions .
Here is onearticle about FFT in IPP and MKL
But both IPP and MKL include convolution functionalities,whichare implementedusing FFT and general convoluation.
Please See ipp manual or MKL manual by visit the Intel MKL web site and the Intel IPP web site.
You may use them directly and see if they can meet your requirement.