Monday, March 2

MS31
Parallel Sparse Matrix Computations and Enabling Algorithms - Part II of II

4:30 PM - 6:30 PM
Room: Symphony IV

For Part I, see MS20

Parallel sparse matrix computations are abundant in computational science and engineering. Sparse matrix factorization algorithms are, perhaps, the most ubiquitous of these computations. Often, graph theoretical algorithms and structures are used to improve the performance and to enable those algorithms.

The current minisymposium gathers the latest developments in the standard (LU and QR) matrix factorizations and in the enabling graph theoretical algorithms and structures, such as partitioning, node ordering, and bipartite graph matching. In doing so, the aim is to emphasize the interaction between these two classes of developments, to facilitate the incubation of new ideas and collaborations, and by inference to enforce the research in both domains.

Organizer: Umit V. Catalyurek
The Ohio State University
Bora Ucar
LIP-ENS Lyon, France
Erik G. Boman
Sandia National Laboratories

4:30-4:55 A New Heuristic for Bipartite Matching Algorithms

Johannes Langguth, University of Bergen, Norway

5:00-5:25 Memory-aware Scheduling for Parallel Out-of-core Multifrontal Factorizations

Emmanuel Agullo, University of Tennessee, Knoxville

5:30-5:55 Sparse Matrix-matrix Multiplication for Accelerating Parallel Graph Computations

Aydin Buluc and John R. Gilbert, University of California, Santa Barbara

6:00-6:25 A Parallel Half-Approximation Algorithm for the Weighted Matching Problem

Mahantesh Halappanavar, Old Dominion University; Florin Dobrian, Columbia University; Alex Pothen, Purdue University