Hash-storage techniques for adaptive multilevel solvers and their domain decomposition parallelization

Griebel, Michael; Zumbusch, Gerhard

doi:10.1090/conm/218/3018

Cited by 23 publications

(18 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also for the efficient and balanced parallelisation of a solver code on adaptively refined grids, space-filling curves have shown to be a very useful tool within the last years [14,35]. For our code, we use this approach, too, and achieve a speedup of 12.93 (strong scaling) for the solution of the poisson equation on an adaptive grid with 23, 118, 848 degrees of freedom run on a myrinet cluster consisting of eight dual 1 In our flow solver, the velocities are vertex data.…”

Section: Grids and Data Structuresmentioning

confidence: 99%

An Eulerian approach for partitioned fluid–structure simulations on Cartesian grids

et al. 2008

View full text Add to dashboard Cite

This paper describes an Eulerian approach for partitioned fluid-structure simulations based on a fluid solver using regularly and adaptively refined Cartesian grids. The particular focus is on the efficient implementation and embedding of the fluid solver in the context of coupled simulations. Special subjects are the efficient layout of data structures and data access based on space-filling curves and the realisation of geometry and topology changes. In addition, a coupling environment is presented that allows for an easy and flexible coupling of flow and structure codes. Simulation results are provided for large particle movements within the drift ratchet scenario.

show abstract

Section: Grids and Data Structuresmentioning

confidence: 99%

An Eulerian approach for partitioned fluid–structure simulations on Cartesian grids

et al. 2008

View full text Add to dashboard Cite

show abstract

“…The best-known example of sequences of lists may be in numerical computing where sparse matrices are represented either as sequences of sparse vectors or as vertices and edges of adjacency or incidence graphs. There are several examples of hash tables used in high-performance computing [36,37]. The Multipol library of distributed data structures from Yelick et al [38] contains a general-purpose non-blocking distributed hash table.…”

Section: A Close Look At a Distributed Hash Table Librarymentioning

confidence: 99%

A role for Pareto optimality in mining performance data

Malard

2004

Concurrency Computat.: Pract. Exper.

View full text Add to dashboard Cite

SUMMARYImprovements in performance modeling and identification of computational regimes within software libraries is a critical first step in developing software libraries that are truly agile with respect to the application as well as to the hardware. It is shown here that Pareto ranking, a concept from multiobjective optimization, can be an effective tool for mining large performance datasets. The approach is illustrated using software performance data gathered using both the public domain LAPACK library and an asynchronous communication library based on IBM LAPI active message library.

show abstract

“…Generalizations of Hilbert curves to higher dimensions are given in [1]. Specific applications include matrix multiplication [23,43], domain decomposition [3,49,86], and image processing [2,4,59,60,77,102]. They are also a standard tool in the creation of cache-oblivious algorithms [13,14,18,44,87,88], which have asymptotically optimal memory performance on multilevel memory hierarchies while avoiding memory-specific parameterization.…”

Section: One-dimensional Reductionmentioning

confidence: 99%

Algorithmic support for commodity-based parallel computing systems.

Leung¹,

Bender

Bunde

et al. 2003

View full text Add to dashboard Cite

Follows 3 AbstractThe Computational Plant or Cplant is a commodity-based distributedmemory supercomputer under development at Sandia National Laboratories. Distributed-memory supercomputers run many parallel programs simultaneously. Users submit their programs to a job queue. When a job is scheduled to run, it is assigned to a set of available processors. Job runtime depends not only on the number of processors but also on the particular set of processors assigned to it. Jobs should be allocated to localized clusters of processors to minimize communication costs and to avoid bandwidth contention caused by overlapping jobs.This report introduces new allocation strategies and performance metrics based on space-filling curves and one dimensional allocation strategies. These algorithms are general and simple. Preliminary simulations and Cplant experiments indicate that both space-filling curves and one-dimensional packing improve processor locality compared to the sorted free list strategy previously used on Cplant. These new allocation strategies are implemented in Release 2.0 of the Cplant System Software that was phased into the Cplant systems at Sandia by May 2002.Experimental results then demonstrated that the average number of communication hops between the processors allocated to a job strongly correlates with the job's completion time. This report also gives processor-allocation algorithms for minimizing the average number of communication hops between the assigned processors for grid architectures.The associated clustering problem is as follows: Given n points in d , find k points that minimize their average pairwise L 1 distance. Exact and approximate algorithms are given for these optimization problems. One of these algorithms has been implemented on Cplant and will be included in Cplant System Software, Version 2.1, to be released. In more preliminary work, we suggest improvements to the scheduler separate from the allocator.

show abstract

Hash-storage techniques for adaptive multilevel solvers and their domain decomposition parallelization

Cited by 23 publications

References 9 publications

An Eulerian approach for partitioned fluid–structure simulations on Cartesian grids

An Eulerian approach for partitioned fluid–structure simulations on Cartesian grids

A role for Pareto optimality in mining performance data

Algorithmic support for commodity-based parallel computing systems.

Contact Info

Product

Resources

About