2026-04-13T06:32:06Zhttps://recercat.cat/oai/request

oai:recercat.cat:2117/1296432025-07-17T05:51:07Zcom_2072_1033col_2072_452950

Strategies for the heterogeneous execution of large-scale simulations on hybrid supercomputers Álvarez Farré, Xavier Gorobets, Andrei Trias Miquel, Francesc Xavier Universitat Politècnica de Catalunya. Departament de Màquines i Motors Tèrmics Universitat Politècnica de Catalunya. CTTC - Centre Tecnològic de la Transferència de Calor Àrees temàtiques de la UPC::Enginyeria mecànica::Mecànica de fluids Àrees temàtiques de la UPC::Informàtica Application software porting Computational fluid dynamics Heterogeneous computing Hybrid computer simulation Parallel CFD SpMV Portability MPI + OpenMP + OpenCL Hybrid CPU + GPU Heterogeneous computing Programari d'aplicació -- Portabilitat Dinàmica de fluids computacional Simulacio per ordinador Massively-parallel devices of various architectures are being adopted by the newest supercomputers to overcome the actual power constraint in the context of the exascale challenge. This progress leads to an increasing hybridisation of HPC systems and makes the design of computing applications a rather complex problem. Therefore, the software e¿ciency and portability are of crucial importance. In this context of accelerated innovation, we developed the HPC2 (Heterogeneous Portable Code for HPC). It is a portable, algebra-based framework for heterogeneous computing with many potential applications in the ¿elds of computational physics and mathematics, such as modelling of incompressible turbulent ¿ows. In its application to CFD, the algorithm of the timeintegration phase relies on a reduced set of only three algebraic operations: the sparse matrix-vector product, the linear combination of vectors and the dot product. This algebraic approach combined with a multilevel MPI+OpenMP+OpenCL parallelisation naturally provides modularity and portability. In this work, we present the strategies for the e¿cient heterogeneous execution of large-scale simulations on hybrid supercomputers that are part of the HPC2 core. The performance is studied in detail for the major computing kernel, the sparse matrix-vector product, using a sparse matrix derived from a simulation on a hybrid unstructured mesh and up to 32 nodes of a hybrid CPU+GPU supercomputer. Postprint (published version) 2018 Conference lecture Alvarez, X.; Gorobets, A.; Trias, F. X. Strategies for the heterogeneous execution of large-scale simulations on hybrid supercomputers. A: European Conference on Computational Mechanics (Solids, Structures and Coupled Problems) / European Conference on Computational Fluid Dynamics. "Proceedings of the 6th European Conference on Computational Mechanics (Solids, Structures and Coupled Problems), ECCM 6, 7th European Conference on Computational Fluid Dynamics, ECFD 7, Glasgow, Scotland, UK, June 11-15, 2018". Barcelona: International Centre for Numerical Methods in Engineering (CIMNE), 2018, p. 2021-2031. 978-84-947311-6-7 https://hdl.handle.net/2117/129643 eng http://www.eccm-ecfd2018.org/admin/files/fileabstract/a2109.pdf info:eu-repo/grantAgreement/MINECO//ENE2014-60577-R/ES/DESARROLLO DE CODIGOS Y ALGORITMOS PARALELOS DE ALTAS PRESTACIONES PARA LA MEJORA DE LA EFICIENCIA EN LOS SECTORES EOLICO, SOLARTERMICO Y EDIFICACION/ info:eu-repo/grantAgreement/MINECO//RYC-2012-11996/ES/RYC-2012-11996/ Restricted access - publisher's policy 11 p. application/pdf International Centre for Numerical Methods in Engineering (CIMNE)