Abstract
The parallel properties of three fast direct solution methods for linear systems with separable block tridiagonal matrices and a related C/MPI code are studied. Fast algorithm for separation of variables and two new variants of the generalized marching algorithm are first summarized. The results from numerical tests performed on two coarse-grained parallel architectures are then reported. The obtained speed-up and efficiency coefficients are compared. The presented results confirm that not always the best sequential solver has the best parallel performance. © Springer-Verlag 2004.
Cite
CITATION STYLE
Bencheva, G. (2004). Parallel performance comparison of three direct separable elliptic solvers. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2907, 421–428. https://doi.org/10.1007/978-3-540-24588-9_48
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
