Computational Technology Resources - CCP

Keywords: kinetic schemes, gas dynamics, explicit difference schemes, stability, high performance computing, hybrid architecture, NUMA technology.

Summary

HPC systems based on multicore processors which have appeared recently require software being created to take into account the hybrid structure of the memory: distributed among the processors and shared among the cores inside the processor. As it appears from experience, it is important for an effective application of multicore processors to apply the algorithms as simple as possible from the logic point of view. In this regard explicit schemes, which can be easily adapted to the architecture of the multicore systems and allow to effectively employ the systems consisting of 10³-10⁴ cores within a single simulation run are very promising. But the explicit schemes impose severe stability limitations on the size of the time step, especially when the parabolic equations are solved. So, the development of explicit schemes with a rather mild stability condition is one of the most important trends.

This paper presents some results of modelling continuous media problems on computer system MVS-Express with a hybrid architecture. Each node of this system contains two AMD Quad-Core Opteron processors and two NVidia CUDA GPU. A finite difference scheme is based on the quasi gas dynamic (QGD) equations system. The successful experience in solving a wide variety of gas dynamic problems by means of QGD based schemes showed that they describe viscous heat conducting flows as good as schemes for Navier-Stokes equations, where the latter are applicable. The explicit scheme, based on physically founded flux relaxation approach described here, has a Courant-like stability condition even for very low Mach numbers and minimum memory requirements per computational point. So, it is very convenient for computer systems with the hybrid architecture, in particular for GPU-based computers. The effective exploitation of such systems is now a serious problem. Parallel realization is based on shmem programming technology intermediate between models for systems with the global memory and systems without it. The calculations results show good parallelization efficiency.

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 25 Numerical Simulation of Continuous Media Problems on Hybrid Computer Systems B.N. Chetverushkin, E.V. Shilnikov and A.A. Davydov Keldysh Institute of Applied Mathematics, Russian Academy of Science, Moscow, Russia doi:10.4203/ccp.95.25 purchase the full-text of this paper Full Bibliographic Reference for this paper B.N. Chetverushkin, E.V. Shilnikov, A.A. Davydov, "Numerical Simulation of Continuous Media Problems on Hybrid Computer Systems", in , (Editors), "Proceedings of the Second International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 25, 2011. doi:10.4203/ccp.95.25 Keywords: kinetic schemes, gas dynamics, explicit difference schemes, stability, high performance computing, hybrid architecture, NUMA technology. Summary HPC systems based on multicore processors which have appeared recently require software being created to take into account the hybrid structure of the memory: distributed among the processors and shared among the cores inside the processor. As it appears from experience, it is important for an effective application of multicore processors to apply the algorithms as simple as possible from the logic point of view. In this regard explicit schemes, which can be easily adapted to the architecture of the multicore systems and allow to effectively employ the systems consisting of 10³-10⁴ cores within a single simulation run are very promising. But the explicit schemes impose severe stability limitations on the size of the time step, especially when the parabolic equations are solved. So, the development of explicit schemes with a rather mild stability condition is one of the most important trends. This paper presents some results of modelling continuous media problems on computer system MVS-Express with a hybrid architecture. Each node of this system contains two AMD Quad-Core Opteron processors and two NVidia CUDA GPU. A finite difference scheme is based on the quasi gas dynamic (QGD) equations system. The successful experience in solving a wide variety of gas dynamic problems by means of QGD based schemes showed that they describe viscous heat conducting flows as good as schemes for Navier-Stokes equations, where the latter are applicable. The explicit scheme, based on physically founded flux relaxation approach described here, has a Courant-like stability condition even for very low Mach numbers and minimum memory requirements per computational point. So, it is very convenient for computer systems with the hybrid architecture, in particular for GPU-based computers. The effective exploitation of such systems is now a serious problem. Parallel realization is based on shmem programming technology intermediate between models for systems with the global memory and systems without it. The calculations results show good parallelization efficiency. 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