ParILUT – A Parallel Threshold ILU for GPUs

TitleParILUT – A Parallel Threshold ILU for GPUs
Publication TypeConference Paper
Year of Publication2019
AuthorsAnzt, H., T. Ribizel, G. Flegar, E. Chow, and J. Dongarra
Conference NameIEEE International Parallel and Distributed Processing Symposium (IPDPS)
Date Published2019-05
PublisherIEEE
Conference LocationRio de Janeiro, Brazil
Abstract

In this paper, we present the first algorithm for computing threshold ILU factorizations on GPU architectures. The proposed ParILUT-GPU algorithm is based on interleaving parallel fixed-point iterations that approximate the incomplete factors for an existing nonzero pattern with a strategy that dynamically adapts the nonzero pattern to the problem characteristics. This requires the efficient selection of thresholds that separate the values to be dropped from the incomplete factors, and we design a novel selection algorithm tailored towards GPUs. All components of the ParILUT-GPU algorithm make heavy use of the features available in the latest NVIDIA GPU generations, and outperform existing multithreaded CPU implementations.

DOI10.1109/IPDPS.2019.00033
Project Tags: 
External Publication Flag: