Solving large systems of linear equations on supercomputers
Mechanical Engineering Associate Professor Dan Negrut will collaborate with colleagues at Purdue University and University of Chicago to produce algorithms capable of solving very large linear algebra problems. Negrut, Ahmed Sameh of Purdue and Matthew Knepley of the University of Chicago received a $600,o00 National Science Foundation grant to investigate how emerging parallel computer hardware can be leveraged to solve systems of linear equations with hundreds of millions of unknowns.
To this end, the team will use what is called heterogeneous computing, that is, combining computing on classical CPUs and on graphical processing unit (GPU) cards. The two classes of problems targeted under this project are banded dense and sparse general linear systems. Experiences and lessons learned in this project will augment a graduate level class, High Performance Computing for Engineering Applications, and the team will share findings at the International Conference for High Performance Computing, Networking, Storage and Analysis and a one-day high performance computing boot camp organized in conjunction with the American Society of Mechanical Engineers conference. In addition, this project will shape the research agendas of two graduate students working on advanced degrees in computational science at Wisconsin and Purdue.