@inproceedings{5c3b9b142ac94db7afd1594a4a5c219e,
title = "FFT blitz: The tensor cores strike back",
abstract = "The fast Fourier Transform (FFT), a reduced-complexity formulation of the Discrete Fourier Transform (DFT), is an important tool in many areas of science and engineering. FFTW is a well-known package that follows this approach and is currently one of the fastest available implementations of the FFT. NVIDIA introduced its version of FFTW called cuFFT that achieves high performance on the GPUs. In this work we present a novel way to map the FFT algorithm on the newly introduced Tensor Cores by adapting the the Cooley-Tukey recursive FFT algorithm. We present four major types of optimizations that enhance the performance of our approach for varying FFT sizes and show that the approach consistently outperforms cuFFT with a speedup of about 15% to 250% on average. ",
keywords = "DFT, FFT, GPU, tensor cores",
author = "Sultan Durrani and Chughtai, {Muhammad Saad} and Abdul Dakkak and Hwu, {Wen Mei} and Lawrence Rauchwerger",
note = "Publisher Copyright: {\textcopyright} 2021 Owner/Author.; 26th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPoPP 2021 ; Conference date: 27-02-2021 Through 03-03-2021",
year = "2021",
month = feb,
day = "17",
doi = "10.1145/3437801.3441623",
language = "English (US)",
series = "Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP",
publisher = "Association for Computing Machinery",
pages = "488--489",
booktitle = "PPoPP 2021 - Proceedings of the 2021 26th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming",
}