Efficient algorithms for computing a rank-revealing UTV factorization on parallel computing architectures

TL;DR

This paper introduces optimized parallel algorithms for the randUTV matrix factorization, enhancing efficiency and scalability on shared and distributed memory systems for low-rank matrix approximations.

Contribution

It presents novel parallel implementations of randUTV, improving speed and scalability over existing methods on high-performance computing architectures.

Findings

randUTV outperforms column pivoted QR in speed on high-performance systems.

Shared memory implementation reduces bottlenecks through task scheduling.

Distributed implementation using ScaLAPACK shows favorable performance.

Abstract

The randomized singular value decomposition (RSVD) is by now a well established technique for efficiently computing an approximate singular value decomposition of a matrix. Building on the ideas that underpin the RSVD, the recently proposed algorithm "randUTV" computes a FULL factorization of a given matrix that provides low-rank approximations with near-optimal error. Because the bulk of randUTV is cast in terms of communication-efficient operations like matrix-matrix multiplication and unpivoted QR factorizations, it is faster than competing rank-revealing factorization methods like column pivoted QR in most high performance computational settings. In this article, optimized randUTV implementations are presented for both shared memory and distributed memory computing environments. For shared memory, randUTV is redesigned in terms of an "algorithm-by-blocks" that, together with a runtime task scheduler, eliminates bottlenecks from data synchronization points to achieve acceleration over the standard "blocked algorithm", based on a purely fork-join approach. The distributed memory implementation is based on the ScaLAPACK library. The performances of our new codes compare favorably with competing factorizations available on both shared memory and distributed memory architectures.