2/11/2016 MechSE Communications
MechSE Adjunct Associate Professor Seid Koric has efficiently solved large real-world and ill-conditioned implicit finite element problems from multiphysics and geophysics with the Watson Sparse Matrix Package (WSMP) solver on the
Written by MechSE Communications
MechSE Adjunct Associate Professor Seid Koric has efficiently solved large real-world and ill-conditioned implicit finite element problems from multiphysics and geophysics with the Watson Sparse Matrix Package (WSMP) solver on the Blue Waters petascale computing system at Illinois' National Center for Supercomputing Applications (NCSA).
A typical finite element code takes many days to solve what Koric and his collaborators solved on the 65,536 cores of Blue Waters in less than a minute.
Solving a linear system of equations accounts for 70-80 percent of the total computational time in many problems in computational science and engineering. When discretizing large multiphysics problems with the implicit finite element method, the associated matrix is sparse, symmetric, and often ill-conditioned. Because of this, these systems are frequently unsolvable by traditional iterative methods.
The main advantages of direct methods are their generality and robustness. But the limitation of CPU computing power and memory requirements has made the use of direct solvers uneconomical, resulting in broad use of iterative solvers.
The unprecedented level of parallel scalability and robustness of direct solvers demonstrated for the first time in this analysis by Koric and his collaborators from IBM Watson and the Barcelona Supercomputing Center could lead to dramatic advances in high-fidelity modeling and design in manufacturing, as well as a better understanding of the properties of the Earth’s subsurface for oil exploration.
Koric’s findings were recently published in two computational mechanics and geophysics journals: Computer Methods in Applied Mechanics and Engineering and Computers & Geosciences.
Koric is also a Technical Program Manager for the Private Sector Program at NCSA, and he joined the MechSE department in 2010. He earned a PhD from Illinois’ mechanical and industrial engineering program in 2006, an MS degree in aerospace engineering at Illinois in 1999, and a BS degree in mechanical engineering from the University of Sarajevo, Bosnia and Herzegovina in 1993.