DARPA/ITO Progress Report for the project ``A scalable library
for numerical linear algebra'' for fiscal year 1995/1996
Project Summary
The objective of the project ``A Scalable Library for Numerical Linear
Algebra'' is to produce a software library for performing dense and
sparse linear algebra computations on massively parallel computers.
These types of computations are used extensively in scientific and
engineering applications, and so the libraries developed will be a key
component in the effective use of massively parallel computers for
performing large-scale scientific computations, such as those targeted
by the High Performance Computing and
Communications (HPCC) program.
The ScaLAPACK project is made up of 4 components:
- dense matrix software (ScaLAPACK)
- large sparse eigenvalue software (PARPACK)
- sparse direct systems software (CAPSS)
- preconditioners for large sparse iterative solvers (PARPRE)
Collaborators
The ScaLAPACK project is a collaborative effort between the following
institutions and principal investigators:
-
University of
Tennessee, Knoxville
- Jack J.
Dongarra, Principal Investigator
PROGRESS REPORT 1995/1996
Jack. J. Dongarra is working on the design and implementation of
ScaLAPACK,
a library of software for doing dense
and banded linear algebra computations on distributed memory concurrent
computers.
-
Mathematical Sciences Section of
Oak Ridge National Laboratory
- David W. Walker, Co-Principal Investigator
PROGRESS REPORT 1995/1996
David W. Walker is also involved in the design and implementation of
ScaLAPACK.
-
Rice University
- Danny Sorensen, Co-Principal Investigator
PROGRESS REPORT 1995/1996
Danny Sorensen is working on the design and implementation of
PARPACK, a distributed-memory library of software that uses the Arnoldi method
for solving large-scale eigenvalue problems for symmetric matrices.
-
University of California, Los Angeles
- Tony Chan, Co-Principal Investigator
PROGRESS REPORT 1995/1996
Tony Chan is doing research into iterative solvers for sparse
nonsymmetric matrices, and is involved in the design and implementation
of PARPRE,
a package of preconditioners for large sparse iterative solvers.
-
University of California, Berkeley
-
James Demmel, Co-Principal Investigator
PROGRESS REPORT 1995/1996
James Demmel is investigating the generalized singular value
decomposition of two matrices, and the use of the matrix sign function
for locating eigenvalues in a specified parallelogram in the complex plane.
He is also involved in the design and implementation of
ScaLAPACK.
-
University of Illinois, Urbana-Champaign
- Michael T. Heath, Co-Principal Investigator
Padma Raghavan, Co-Principal Investigator
PROGRESS REPORT 1995/1996
Michael T. Heath and Padma Raghavan are developing software for the symbolic
and numeric Cholesky factorization of symmetric, positive-definite matrices.
They are
involved in the design and implementation of the CAPSS project.
The Scalable Libraries Project is largely funded by the
Defense Advanced Research Projects Agency.
DARPA's Information Technology Office sponsors a variety of Research Programs.
Click here to return
to the index for the ScaLAPACK project.
Last Modified July 15, 1996