FFT Speed up Generation

FFT Speed up Generation

Hey there

I d like to use the MKL library to generate the 1D FFT(complex to complex), with the size of 256.

To speed up the process i want to split the 256 FFT vector to 4 sub-vectors of 64 and run each on 1 thread (i have 4 threads on my processor).

How with the help of OMP library i can get all the sub-vectors executed simultaneously in the parallel mode? Any help or programming details will be apperciated.

Thanks

3 posts / 0 nouveau(x)
Dernière contribution
Reportez-vous à notre Notice d'optimisation pour plus d'informations sur les choix et l'optimisation des performances dans les produits logiciels Intel.

Please take a look at a thread http://software.intel.com/en-us/forums/topic/402439 for more technical details on how to use MKL's DFTI functions. Then, a source data set of 256 elements is too small and I think in the end an overhead of 4-thread processing won't give you any performance gains. However, it will be a great academic experience for you on how to do some processing in parallel.

Regarding OpenMP programming take a look at http://openmp.org/wp. Resources web-page has a link to Tutorials and Articles.

Thank you for the reply

Connectez-vous pour laisser un commentaire.