Special issue on Parallel Matrix Algorithms and Applications

Abstract

This special issue of Parallel Computing contains nine articles, selected after peer reviewing, from invited and contributed presentations made at the 8th International Workshop on Parallel Matrix Algorithms and Applications (PMAA'16), that took place at the Universite of Bordeaux, France, from July 6-8, 2016. The workshop attracted around 120 participants from all continents, 25% were PhD students and around 10% from industry. The workshop was co-chaired by Emmanuel Agullo, Peter Arbenz, Luc Gi-raud, and Olaf Schenk. The members of the program committee were : P. D'Am-bra, H A total of twelve high quality submissions were received. In this special issue nine eventually accepted papers appear. The nine papers address diverse aspects of linear algebra and high performance computing 1. Jack Dongarra, Mark Gates, Stanimire Tomov address accelerating the SVD two stage reduction and divide-and-conquer using GPUs. The increasing gap between memory bandwidth and computation speed motivates the choice of algorithms to take full advantage of today's high performance computers. For dense matrices, the classic algorithm for the SVD uses a one-stage reduction to bidiagonal form, which is limited in performance by the memory bandwidth. To overcome this limitation, a two-stage reduction to bidiagonal has been gaining popularity. As accelerators , such as GPUs and co-processors, are becoming increasingly widespread in high-performance computing, the authors present an accelerated SVD employing a two-stage reduction to bidiagonal as well as a parallelized and accelerated divide-and-conquer algorithm to solve the subsequent bidiagonal SVD. The new implementation provides a significant speedup compared to existing multi-core and GPU-based SVD implementations.