Next-generation FFT Algorithms at PP24
The 2024 SIAM Conference on Parallel Processing for Scientific Computing (PP24), which took place in Baltimore, Md., earlier this month, featured a minisymposium on “Next Generation FFT Algorithms in Theory and Practice: Parallel Implementations and Applications.” This session brought together a variety of researchers who are studying cutting-edge fast Fourier transform (FFT) algorithms and their parallel implementations.
Daisuke Takahashi (University of Tsukuba in Japan), Franz Franchetti (Carnegie Mellon University), and I organized the minisymposium, which featured four presentations about various aspects of next-generation FFT algorithms. Takahashi opened the session with a talk titled “Parallel Implementations of Number Theoretic Transforms (NTTs) on GPU Clusters.” His remarks introduced new approaches that utilize the power of graphics processing unit (GPU) clusters to perform NTTs, which are closely related to FFTs. Next, Sanil Rao (Carnegie Mellon University) spoke about “FFTX: Release, Updates, and Next Steps” and provided an update on the FFTX library: a software package for the development of high-performance FFT algorithms.
During my presentation on the “Comparison of Benchmarks for Next Generation FFT Algorithms,” I compared the performance of Intel and OSU all-to-all communication benchmarks to glean valuable insights about communication patterns that are relevant to FFT functionalities on modern hardware architectures. Finally, Miguel Ferrer Avila (NVIDIA) closed out the session with his talk about the “Latest Advancements in Parallel and Distributed FFT Computation on NVIDIA GPUs,” which explored recent updates surrounding the use of NVIDIA GPUs for FFT computations.
The two-hour PP24 session, which focused on the future directions of FFT algorithms and their practical applications, provided a valuable overview of the current state of the art and fostered a lively discussion between researchers from both academia and industry. If you are interested in learning more about this topic, please visit the FFT Report website for access to slides from these and other presentations from previous conferences.
About the Author
Samar Aseeri
Computational scientist, King Abdullah University
Samar Aseeri is a computational scientist at King Abdullah University of Science and Technology (KAUST) in Saudi Arabia. She earned her undergraduate, graduate, and doctoral degrees in applied mathematics from Umm Al-Qura University in Saudi Arabia, and completed supercomputing training from IBM in New York. Aseeri is currently leading two initiatives to establish high-performance computing communities: Benchmarking in the Data Center and FFT in the Exascale Era.