Abstracts: Jennifer Scott
Too large to handle directly?
A key strength of the mathematical software library HSL and one for which it is internationally renowned is its collection of packages for the direct solution of sparse linear systems of equations. Direct methods are popular for their robustness and efficiency. Their main weakness is that the memory they require usually increases rapidly with problem size. One way of attempting to overcome this limitation is to use a solver that is able to hold its data structures on disk, that is, an out-of-core solver. John Reid has long been interested in out-of-core solvers; indeed, twenty five years ago he wrote an out-of-core multifrontal solver for finite-element systems. As the requirements of computational scientists for more accurate models increases, so inevitably do the sizes of the systems that must be solved and this has led in recent years to renewed interest in out-of-core solvers.
In this talk, we briefly review the development of out-of-core solvers within HSL and then discuss the design of our latest out-of-core solver, HSL_MA77. HSL_MA77 implements a multifrontal out-of-core algorithm. It relies on direct-access files to hold the original matrix, its factorization, and the bulk of its intermediate data. We highlight some of the key features of the new package and present numerical results for a range of large-scale problems arising from practical applications.
This is joint work with John Reid.