https://community.intel.com/t5/Intel-Integrated-Performance/Fast-vs-Accurate-FFT/m-p/925198#M16108 <P>Hi Nimrod,</P> <P>IPP FFT uses Cooley-Tukey algorithm. Initially, the IppHintAlgorithm argument allowed to choose whether FPU (ippAlgHintAccurate) or MMX (ippAlgHintFast) implementation should be used. Since IPP takes advantage of SSE, AVX and AVX2 instructions, now this option is no longer supported and exists only for API compatibility with previous versions of IPP. Thank you for noticing this - we will update the documentation.</P> <P>Regards, Alexey</P> Fri, 28 Feb 2014 15:29:47 GMT Alexey_Tyndyuk 2014-02-28T15:29:47Z https://community.intel.com/t5/Intel-Integrated-Performance/Fast-vs-Accurate-FFT/m-p/925197#M16107 <P>Hi ,</P> <P>What is the difference between the fast (ippAlgHintFast) and Accurate (ippAlgHintAccurate) FFTs ? Which algorithm each one of the options use ?</P> <P>Thanks , Nimrod</P> Tue, 25 Feb 2014 15:40:12 GMT https://community.intel.com/t5/Intel-Integrated-Performance/Fast-vs-Accurate-FFT/m-p/925197#M16107 Nimrod_H_ 2014-02-25T15:40:12Z https://community.intel.com/t5/Intel-Integrated-Performance/Fast-vs-Accurate-FFT/m-p/925198#M16108 <P>Hi Nimrod,</P> <P>IPP FFT uses Cooley-Tukey algorithm. Initially, the IppHintAlgorithm argument allowed to choose whether FPU (ippAlgHintAccurate) or MMX (ippAlgHintFast) implementation should be used. Since IPP takes advantage of SSE, AVX and AVX2 instructions, now this option is no longer supported and exists only for API compatibility with previous versions of IPP. Thank you for noticing this - we will update the documentation.</P> <P>Regards, Alexey</P> Fri, 28 Feb 2014 15:29:47 GMT https://community.intel.com/t5/Intel-Integrated-Performance/Fast-vs-Accurate-FFT/m-p/925198#M16108 Alexey_Tyndyuk 2014-02-28T15:29:47Z