this paper presents an experimental evaluation of a suite of dynamic partitioning/load-balancing techniques for adaptive grid hierarchies that underlie parallel structured adaptive mesh refinement applications. Partit...
详细信息
Withthe availability of Computational Grids, new kinds of applications that will soon emerge will raise the problem of how to program them on such computing systems. In this paper, we advocate a programming model tha...
详细信息
A parallel single level store (psls) system integrates a shared virtual memory and a parallel file system representing an attractive support for long running parallel applications in a cluster. In this paper we presen...
详细信息
Porous laminar materials and alternate laminates of dense and porous layers in silicon carbide have been elaborated by tape casting and liquid phase sintering (YAG-alumina eutectic) processing. Porosity was introduced...
详细信息
Porous laminar materials and alternate laminates of dense and porous layers in silicon carbide have been elaborated by tape casting and liquid phase sintering (YAG-alumina eutectic) processing. Porosity was introduced by the incorporation of pore forming agents (PFA) (5-50 vol%) in the slurry. Two types of PFA with a narrow size distribution have been used (corn starch and polyamide powders). the effects of size, content, type of PFA on tape casting processing, sintering characteristics, porosity control and microstructure are investigated. For each PFA, the porosity attains a maximum value dependent on the PFA nature (41 vol% with corn starch). Only for starch, the volumetric shrinkage was unaffected by the PFA content up to 45 vol%. Homogeneous distribution of porosity has been obtained for both monolithic and composite laminates. An equiaxed and homogeneous silicon carbide microstructure has been obtained and was unaffected by PFA. Layered structures without defects have been obtained withparallel layers and uniform thickness (dense layer: 70 mum;porous layer: 80 mum). (C) 2001 Published by Elsevier Science Ltd.
Current analytic solutions to the execution time prediction Y of binary parallel compositions of tasks with arbitrary execution time distributions X1 and X2 are either computationally complex or very inaccurate. In th...
详细信息
Multipartitioning is a skewed-cyclic block distribution that yields better parallel efficiency and scalability for line-sweep computations than traditionalb lock partitionings. this paper describes extensions to the R...
详细信息
We describe the design and the implementation in Java and Jini of a Computational Community, which supports the federation of resources from different organisations. Resources from the local Administrative Domain are ...
详细信息
the MMX and SSE extensions of current Intel Pentium processors offer a 4-way or 8-way SIMD parallelism to accelerate many vector or matrix applications. In this paper the performance of MMX and SSE for the implementat...
详细信息
We describe and analyze the parallel implementation of a novel domain decomposition preconditioner for the fast iterative solution of linear systems of algebraic equations arising from the discretization of elliptic p...
详细信息
An integrated storage platform for open systems should be able of meeting the requirements of deterministic applications, multimedia systems, and traditional best-effort applications. It should also provide a disk sch...
详细信息
ISBN:
(纸本)3540419993
An integrated storage platform for open systems should be able of meeting the requirements of deterministic applications, multimedia systems, and traditional best-effort applications. It should also provide a disk scheduling mechanism fitting all those types of applications. In this paper, we propose a three-level hieraxchical. disk scheduling scheme, which has three main components. metascheduler, single server scheduler, and disk scheduler. the metascheduler provides scheduling mechanisms for a parallel disk system or a set of parallel servers. the server level is divided in three main queues. deterministic, statistic and best-effort requests. Each server may have its own scheduling algorithm. the lower level, disk driver, chooses the ready streams using its own scheduling criteria. those systems have been implemented and tested, and the performance evaluations demonstrate that our scheduling architecture is adequate for handling stream sets with different timing and bandwidth requirements.
暂无评论