cuFINUFFT: a load-balanced GPU library for general-purpose nonuniform FFTs

TL;DR

This paper introduces cuFINUFFT, a GPU library for nonuniform FFTs that achieves high performance and load balancing, significantly outperforming CPU and existing GPU implementations in various applications.

Contribution

The paper presents a general-purpose, load-balanced GPU library for nonuniform FFTs in 2D and 3D, with high accuracy and superior performance over CPU and existing GPU codes.

Findings

Achieves 10^9 points/sec at low accuracy on GPU.

Up to 90x faster than GPU codes at high accuracy.

Demonstrates 5-12x speedup over CPU in 3D X-ray diffraction reconstruction.

Abstract

Nonuniform fast Fourier transforms dominate the computational cost in many applications including image reconstruction and signal processing. We thus present a general-purpose GPU-based CUDA library for type 1 (nonuniform to uniform) and type 2 (uniform to nonuniform) transforms in dimensions 2 and 3, in single or double precision. It achieves high performance for a given user-requested accuracy, regardless of the distribution of nonuniform points, via cache-aware point reordering, and load-balanced blocked spreading in shared memory. At low accuracies, this gives on-GPU throughputs around $10^9$ nonuniform points per second, and (even including host-device transfer) is typically 4-10$\times$ faster than the latest parallel CPU code FINUFFT (at 28 threads). It is competitive with two established GPU codes, being up to 90$\times$ faster at high accuracy and/or type 1 clustered point distributions. Finally we demonstrate a 5-12$\times$ speedup versus CPU in an X-ray diffraction 3D iterative reconstruction task at $10^{-12}$ accuracy, observing excellent multi-GPU weak scaling up to one rank per GPU.