holds the Regents Chair in Computer Science.
His interests span almost all areas of computer science.
More recently, he has pursued research in
Computational Science, Parallel Computing, and Grid Computing.
Prof. Browne holds joint appointments in the departments of
Computer Sciences, Physics, and Electrical and Computer Engineering.
holds the David Bruton, Jr. Professorship in
Computer Sciences and is also Professor of Mathematics. He
constructed FITPACK, a large package of curve and surface fitting
subprograms that employed tension splines. His interests in fitting
extend to the related areas of approximation theory, grid
construction, and computational geometry. He has done research in
numerical linear algebra - especially condition number estimation and
the theory and use of the singular value decomposition. More recently
he has considered the automatic detection of instability in scientific
software.
pursues research
in computational linear algebra,
data mining
and bioinformatics. His emphasis is on exploring core problems in these
areas to obtain novel algorithms that preserve the underlying problem
structure. Some of these problems include
clustering of high-dimensional data, low-dimensional approximations
that preserve sparsity and non-negativity, and fast algorithms for
eigenvalue problems.
is interested in algorithms, hardware architectures,
and programming environments for 3D graphics systems and
general-purpose single-chip parallel computers.
For scientific computation problems, these future single-chip
parallel computers hold the promise of much higher performance, but
realizing this high performance will likely require significant
changes to scientific computation techniques.
is currently interested in the software engineering aspects of
high-performance sequential and parallel linear algebra libraries.
Current projects include
The
Formal Linear Algebra Methods Environment (FLAME)
project pursues the formal derivation of high-performance linear algebra
algorithms and their implementation. This research has produced a
systematic approach to the derivation of provably correct algorithms.
In conjunction with the FLAME API, these correct algorithms can be
easily translated into correct code. It is the goal of the project
to ultimately facilitate the automation of the derivation
of algorithms, the generation of the code, and the complexity and
stability analysis of the algorithms.
The
GOTO Basic Linear Algebra Subprograms
is a set of high-performance
implementations of the Basic Linear Algebra Subprograms (BLAS)
for various architectures.
The
Parallel Linear Algebra Package (PLAPACK)
project provides an infrastructure for the parallel implementation of
high-performance dense and banded linear algebra operations.
Kathryn McKinley's primary research goal is to enable programmers to
express their applications in high level languages, and to develop
advanced compiler
techniques that enable them to achieve high performance. She is
particularly interested
in improving data locality, and is current pursuing hardware/software
cooperative
caching and prefetching to close the memory gap for scientific and
general purpose
applications.
The
Automatic Theorem Proving Group
in the department has a very
active research project related to the formal verification
of floating point units.
Faculty involved include