Scientific Computing (SCI)

Overview

The faculty, students, visitors, and affiliated researchers of Berkeley's Scientific Computing and Numerical Methods group have produced some of the most heavily used scientific software (and hardware standards!) in the world, backed up with strong theoretical foundations. Their strength is a result of their commitment to building industrial-strength software and algorithms and supporting full-scale scientific applications in large interdisciplinary collaborations among scientists, engineers, mathematicians, and computer scientists. Their research encompasses symbolic, numerical, and geometric computation, often on parallel or distributed systems.

Berkeley is ideally situated amongst a wealth of resources for scientific computing, including Lawrence Berkeley National Laboratory (LBNL), the National Energy Research Scientific Computing Center (NERSC, located at LBNL), and the Lawrence Livermore National Laboratory. Each of these is a source of scientific collaborators, difficult problems, and high-performance computing resources. Other important computational resources are Berkeley's own campus-wide Millenium project, built upon a cluster of clusters of tightly-coupled Intel-based computers. (One of the first clusters ever constructed was the Berkeley Network of Workstations (NOW) project.)

Major contributions from the group include the IEEE floating point standard; LAPACK, ScaLAPACK and SuperLU for numerical linear algebra, the programming languages UPC, Split-C and Titanium, coordination languages, the Triangle mesh generation program, numerical routines for UNIX; and MACSYMA.

Topics

  • Parallel computing

    Languages and numerical algorithms for parallel computers

  • Automatic Performance Tuning

    Automatic generation of optimized implementations of computational and communication kernels, tuned for particular architectures and work loads.

  • Mesh generation

    Automatic generation of triangulated meshes to represent physical and computational domains.

  • Matrix computations

    Numerical algorithms and software for fast and accurate numerical linear algebra.

  • Floating point

    Extended precision arithmetic. Reliable floating point standards. Architectural and run time implications of floating point standards. Programming language implications of floating point standards.

  • Animation

  • Computer Algebra

    Methods for symbolic mathematical computation.

Research Centers

Faculty

Primary

Secondary

Faculty Awards

  • ACM A.M. Turing Award: William M. Kahan, 1989.
  • SIAM John von Neumann Lecture Prize: William M. Kahan, 1997.
  • National Academy of Sciences (NAS) Member: James Demmel, 2011.
  • National Academy of Engineering (NAE) Member: Katherine A. Yelick, 2017. William M. Kahan, 2005. James Demmel, 1999. Susan L. Graham, 1993.
  • American Academy of Arts and Sciences Member: James Demmel, 2018. Katherine A. Yelick, 2017. William M. Kahan, 2003. Susan L. Graham, 1995.
  • Berkeley Citation: Susan L. Graham, 2009.
  • Sloan Research Fellow: Michael Lustig, 2013. Jonathan Shewchuk, 2004. James O'Brien, 2003.

Related Courses