Research Laboratory - High-Performance Parallel Computing Laboratory

The laboratory for high-performance computing focuses mainly on the interaction of computer architecture and numerically intensive algorithms. The lab conducts research on parallel algorithms, on algorithms that effectively exploit cache-memories, and on algorithms that manipulate disk-resident data structures that are too large to fit in memory. The research in the lab is both theoretical and applied: nearly every project involves design and analysis of algorithms, design and implementation of these algorithms, and extensive experimental evaluations.

The laboratory also focuses on discrete and combinatorial issues in scientific simulations, a research area that lies on the boundary between computer science and applied math. Recent projects in the lab include research in the following areas:

• Communication-efficient parallel algorithms for numerical linear algebra

• Parallel and cache-efficient direct solvers for sparse linear systems

• Out-of-core sparse linear solvers and out-of-core orthogonalization

• Parallel game-playing algorithms (e.g. for Chess)

• Graph algorithms for reordering matrices to reduce fill in Gaussian elimination

• Graph algorithms for designing iterative linear solvers

The lab is supervised by: Prof. Sivan Toledo