FoSSI : The family of simplified solver interfaces for the rapid development of parallel numerical atmosphere and ocean models

Edit Item Edit Item

General Information:

Frickenhaus, S. , Hiller, W. and Best, M. (2005): FoSSI : The family of simplified solver interfaces for the rapid development of parallel numerical atmosphere and ocean models , Ocean Modelling, 10, 1-2, 185-191 - The Second International Workshop on Unstructured Mesh Numerical Modelling of Coastal, Shelf and Ocean Flows. . doi: 10.1016/j.ocemod.2004.06.004
Cite this page as:
Official URL:
Contact Email:

Supplementary Information:


The portable software system FoSSI is presented that allows for an efficient and scalable parallel solution of largesparse linear equations systems arising in finite element methods. The software offers an efficient and easy, yet flexible interface to several parallel solvers available on the web, such as PETSC, AZTEC, MUMPS, PILUT and HYPRE. FoSSI makes use of the concept of handles for vectors, matrices, preconditioners and solvers, that is frequently used in solver libraries. Hence, FoSSI allows for a flexible treatment of several linear equations systems and associated preconditioners at the same time, even in parallel on separate MPI-communicators. The second special feature in FoSSI ist the task specifier, being a combination ofkeywords, each configuring a certain phase in the solver setup. This enables the userto control a solver over one unique subroutine. Furthermore,FoSSI has rather similar features for all solvers, making a fast solver intercomparison or exchange an easy task.FoSSI and its followers is a community software, proven in an adaptive 2D-atmospheremodel and a 3D-primitive equation ocean model.Performance measurements are given for test matrices from these models, demonstratinga minimal yet scalable overhead of the interface.Speedup experiments and solver optimization are reported, giving a brief overview over some of the solver packages performance.It is shown that FoSSI is easily and efficiently used even on LINUX-clusters. Furthermore, an OpenMP-implementation of parallel iterative solvers based on domain dcomposition methods is discussed. This approach to OpenMP solvers is rather attractive, as thecode for domain-local operations of factorization, preconditioning and matrix vector product can be readily taken from a sequential implementation that is also suitable tobe used in the MPI-variant.

Further Details:

read more
OAI 2.0:
ePIC is powered by:
EPrints 3