Computational Technology Resources - CCP

P. Vezolle¹, Y. Fournier², C. Moulinec³ and N. Tallet¹

¹IBM France, Montpellier, France
²EDF R&D, MFEE, Chatou, France
³STFC Daresbury Laboratory, Warrington, United Kingdom

doi:10.4203/ccp.95.65

P. Vezolle, Y. Fournier, C. Moulinec, N. Tallet, "Multithreads and MPI Adaptive AMG Linear Solver for Parallel Navier-Stokes Approaches", in , (Editors), "Proceedings of the Second International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 65, 2011. doi:10.4203/ccp.95.65

Keywords: high-performance computing, parallelisation, domain decomposition, multigrid, multithreads, sparse linear solver, mesh renumbering, Petaflops.

Summary

The paper deals with two parallel versions of an algebraic multigrid solver: a standard MPI version where the coarse grids are solved by all the MPI processes and the adaptive version with a `rank merging' step where coarse grids are gathered from several processes onto one process, by groups of 2-8 processes every few coarse mesh levels. For the muli-threaded approached we outline here how to introduce new level of parallelisms using compiler directives to simultaneously take advantage of the emerging processor features such as many-core, simultaneous multithreading (SMT), transactional memory (TM), and thread level speculation (TLS). In a Finite Volume cell-centered formulation the main loops are face-based. We introduce OpenMP implementattion of the face-based loops and assess how to optimize thread generation with TM and TLS.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £85 +P&P)

	Computational & Technology Resources an online resource for computational, engineering & technology publications
	not logged in - login
Front Page Browse CCP CSETS CTR IJRT Other Authors Search Purchase Guide FAQ Contact us	Civil-Comp Proceedings ISSN 1759-3433 CCP: 95 PROCEEDINGS OF THE SECOND INTERNATIONAL CONFERENCE ON PARALLEL, DISTRIBUTED, GRID AND CLOUD COMPUTING FOR ENGINEERING Edited by: Paper 65 Multithreads and MPI Adaptive AMG Linear Solver for Parallel Navier-Stokes Approaches P. Vezolle¹, Y. Fournier², C. Moulinec³ and N. Tallet¹ ¹IBM France, Montpellier, France ²EDF R&D, MFEE, Chatou, France ³STFC Daresbury Laboratory, Warrington, United Kingdom doi:10.4203/ccp.95.65 purchase the full-text of this paper Full Bibliographic Reference for this paper P. Vezolle, Y. Fournier, C. Moulinec, N. Tallet, "Multithreads and MPI Adaptive AMG Linear Solver for Parallel Navier-Stokes Approaches", in , (Editors), "Proceedings of the Second International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 65, 2011. doi:10.4203/ccp.95.65 Keywords: high-performance computing, parallelisation, domain decomposition, multigrid, multithreads, sparse linear solver, mesh renumbering, Petaflops. Summary The paper deals with two parallel versions of an algebraic multigrid solver: a standard MPI version where the coarse grids are solved by all the MPI processes and the adaptive version with a `rank merging' step where coarse grids are gathered from several processes onto one process, by groups of 2-8 processes every few coarse mesh levels. For the muli-threaded approached we outline here how to introduce new level of parallelisms using compiler directives to simultaneously take advantage of the emerging processor features such as many-core, simultaneous multithreading (SMT), transactional memory (TM), and thread level speculation (TLS). In a Finite Volume cell-centered formulation the main loops are face-based. We introduce OpenMP implementattion of the face-based loops and assess how to optimize thread generation with TM and TLS. purchase the full-text of this paper (price £20) go to the previous paper go to the next paper return to the table of contents return to the book description purchase this book (price £85 +P&P)
Back to top	©Civil-Comp Limited 2023 - terms & conditions