Leveraging task-parallelism in message-passing dense matrix factorizations using SMPSs

In this paper, we investigate how to exploit task-parallelism during the execution of the Cholesky factorization on clusters of multicore processors with the SMPSs programming model. Our analysis reveals that the major difficulties in adapting the code for this operation in ScaLAPACK to SMPSs lie in algorithmic restrictions and the semantics of the SMPSs programming model, but also that they both can be overcome with a limited programming effort. The experimental results report considerable gains in performance and scalability of the routine parallelized with SMPSs when compared with conventional approaches to execute the original ScaLAPACK implementation in parallel as well as two recent message-passing routines for this operation. In summary, our study opens the door to the possibility of reusing message-passing legacy codes/libraries for linear algebra, by introducing up-to-date techniques like dynamic out-of-order scheduling that significantly upgrade their performance, while avoiding a costly rewrite/reimplementation. © 2014 Elsevier B.V. All rights reserved.This research was supported by Project EU INFRA-2010-1.2.2 ‘‘TEXT: Towards EXaflop applicaTions’’. The researcher at BSC-CNS was supported by the HiPEAC-2 Network of Excellence (FP7/ICT 217068), the Spanish Ministry of Education (CICYT TIN2011-23283, TIN2007-60625 and CSD2007-00050), and theGeneralitat de Catalunya (2009-SGR-980). The researcher at CIMNE was partially funded by the UPC postdoctoral grants under the programme ‘‘BKC5-Atracció i Fidelització de talent al BKC’’. The researcher at UJI was supported by project CICYT TIN2008-06570-C04-01 and FEDER. We thank Jesus Labarta, from BSC-CNS, for helpful discussions on SMPSs and his help with the performance analysis of the codes with Paraver . We thank Vladimir Marjanovic, also from BSC-CNS, for his help in the set-up and tuning of the MPI/SMPSs tools on JUROPA . Finally, we thank Rafael Mayo, from UJI, for his support in the preliminary stages of this work. The authors gratefully acknowledge the computing time granted on the supercomputer JUROPA at Jülich Supercomputing Centre.Peer Reviewe