Shared Virtual Memory and Message Passing Programming on a Finite Element Application
نویسندگان
چکیده
This paper describes the methods used and experiences made with implementing a nite element application on three diierent parallel computers with either message passing or shared virtual memory as the programming model. Designing a parallel nite element application using message-passing requires to nd a data domain decomposition to map data into the local memory of the processors. Since data accesses may be very irregular, communication patterns are unknown prior to the parallel execution and thus makes the parallelization a diicult task. We argue that the use of a shared virtual memory greatly sim-pliies the parallelization step. It is shown experimentally on an hypercube iPSC/2 that the use of the KOAN/Fortran-S programming environment based on a shared virtual memory allows to port quickly and easily a sequential application without a signiicant degradation in performance compared to the message passing version. Results for recent parallel archi-tectures such as the Paragon XP/S for message-passing and the KSR1 for shared virtual memory are presented, too. Comparaison des techniques de m emoire virtuelle partag ee et d' echange de messages a l'aide d'une application par el ement nie R esum e : Ce papier pr esente les r esultats d'une mise en oeuvre d'une application par el ement nie sur trois architectures parall eles en utilisant soit l' echange de messages soit une m emoire virtuelle partag ee. La conception d'une application par el ement nie sur une architecture parall ele a m emoire distribu ee n ecessite une d ecomposition de domaine aan de placer les donn ees dans les m emoires locales des processeurs. Les acc es aux donn ees etant irr eguliers, la parall elisation est une t^ ache complexe. Nous montrons que l'utilisation d'une m emoire virtuelle partag ee simpliie grandement cette t^ ache. Nous montrons, a l'aide de r esultats exp erimentaux sur un hypercube iPSC/2, que l'utilisation de l'environnement de programmation KOAN/Fortran-S permet le portage rapide d'une telle application sans constater une perte signiicative. Nous pr esentons egalement des r esultats sur des machines plus r ecentes comme le Paragon XP/S et la KSR-1.
منابع مشابه
A Comparison of Shared Virtual Memory and Message Passing Programming Techniques Based on a Finite Element Application
This paper describes the methods used and experiences made with implementing a nite element application on three diierent parallel computers with either message passing or shared virtual memory as the programming model. Designing a parallel nite element application using message-passing requires to nd a data domain decomposition to map data into the local memory of the processors. Since data ac...
متن کاملPassing Programming on a Finite Element Application
This paper describes the methods used and experiences made with implementing a nite element application on three diierent parallel computers with either message passing or shared virtual memory as the programming model. Designing a parallel nite element application using message-passing requires to nd a data domain decomposition to map data into the local memory of the processors. Since data ac...
متن کاملPERFORMANCE OF HYBRID OpenMP/MPI PARALLEL PROGRAMMING APPLICATION OF FINITE ELEMENT METHOD
In the area of parallel processing and parallel algorithms several processors are used together to execute a single application faster. There are two major programming paradigms: Shared Memory and Message Passing. Each of them fits into a specific physical model, but there are multiprocessors architectures whose mapping to one of these paradigms is not so simple. SMP clusters, for example, are ...
متن کاملMultilevel Parallel Programming for Three-Dimensional Voronoi Cell Finite Element Modeling of Heterogeneous Materials
In this paper we present work aimed at implementing an efficient parallel model for the three-dimensional Voronoi cell finite element model (VCFEM), which has been developed as an effective tool for modeling heterogeneous microstructures of composite, porous or polycrystalline materials, on SMP cluster systems. In these systems, multiprocessor nodes offer shared memory on the node and distribut...
متن کاملParallel Array Classes and Lightweight Sharing Mechanisms
We discuss a set of parallel array classes, MetaMP, for distributed-memory architectures. The classes are implemented in C++ and interface to the PVM or Intel NX message-passing systems. An array. class implements a partitioned array as a set of objects distributed across the nodes a *collectiven object. Object methods hide the low-level message-passing and implement meaningful array operations...
متن کامل