Parallelisation of BICG-STAB algorithm for finite element computations
Links to Fileshttps://userpages.umbc.edu/~sousedik/papers/hejnice03.pdf
MetadataShow full item record
Type of Work11 pages
conference papers and proceedings
Citation of Original PublicationBedřich Sousedík, Jaroslav Novotný: Parallelization of the BICG-STAB algorithm for finite element computations, Proceedings of Software and Algorithms of Numerical Mathematics (15.), Hejnice, Czech Republic, September 8-12, 2003, Institute of Thermomechanics AS CR
RightsThis item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
We describe parallelization and implementation of BICG-STAB method for systems of unsymmetric linear equations arising in nite element discretization. Global matrix is not constructed, we use element by element approach. All element matrices together with addressing vectors are distributed among processors and read into memory. Then multiplication of element matrices by a global vector is performed simultaneously on different processors. Two odes were developed: using MPI library for distributed memory systems and using OpenMP for shared memory systems. Good scalability is demonstrated on an example with about 30000 unknowns.