Parallel implementation of Fast Fourier Transform in Spherical

Parallel implementation of Fast Fourier Transform in Spherical Coordinates Mentor: Dr. Luis Cueva-Parra Intern: Adrian Rodriguez

Outline • • • My project How it is done today My new approach Impact Timeline Benchmarks

The Project Description What I am trying to do • Fast Fourier Transforms are computed every day all around us for signal processing , Medical imaging. Also applied to compute mathematics in Quantum Physics • I will be attempting to parallelize the Fast Fourier Transform algorithm and have it perform the computation corresponding to spherical coordinates, as opposed to cartesian.

How it is done today • FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. • FFTW does have a working library for multidimensional FFT’s for shared memory, and distributed memory.

My approach • I won’t be redeveloping the library • I will be creating a subroutine to run FFT’s using a hybrid approach by implementing MPI, and Open. MP • I will also create a subroutine to implement FFT’s in spherical coordinates �It may be easier to compute.

Impact • Improving efficiency in electrical systems by predicting the electron flow in different mediums.

Timeline Calculus • Gain the mathematics background HPC Important: Document everything • Review Open. MP and MPI Design • Design the algorithm to implement a hybrid of MPI and open. MP

Timeline Implement • Build an executable product Testing • Execute, and debug Maintenance • Continue to debug

Benchmark • What is success: ▫ Code executes arithmetic correctly ▫ Code runs faster in parallel than in serial ▫ Met required time frames • How will I measure it? ▫ Analyze output data with known outputs ▫ get_wtime to output run time. Compare to serial version ▫ Finish

• Thank you