Artifact Details


Dongarra, Jack J. SIAM oral history

Catalog Number





Jack J. Dongarra describes his professional career to date, with particular reference to his involvement in the production of mathematical software packages. He grew up in Chicago and studied mathematics at Chicago State College, where an internship at Argonne National Laboratory involved him in the ongoing EISPACK project and fostered a life long interest in mathematical software. Dongarra then earned a graduate degree in computer science from the Illinois Institute of Technology, while continuing to work with Argonne staff including Brian Smith, Jim Cody, Danny Sorensen, Jim Boyle, and Jim Pool. Dongarra discuses the organization and functioning of the EISPACK group, the mechanisms used to produce, distribute and maintain code, and his personal role in producing test programs. After graduating in 1975 he went to work full time at Argonne, developing mathematical software. During this period he also studied for a Ph.D. at the University of New Mexico under the direction of Cleve Moler, and worked as a visiting scientist at Los Alamos. From 1976 to 1979 he worked with Moler, Pete Stewart and Jim Bunch to produce LINPACK, a widely used collection of routines for linear algebra. LINPACK also became important as a benchmark for the performance of scientific computers. Dongarra discusses the origins, purpose, organization and results of this project. LINPACK was closely tied to BLAS, the Basic Linear Algebra Subprograms. He also explores its parallels with, and differences from, the later open source software movement. Dongarra received his PhD in 1980, and remained at Argonne, where he developed the NETLIB online software repository (in collaboration with Eric Gross of Bell Labs). In 1989 he accepted a joint position between the University of Tennessee and Oak Ridge National Laboratory. Dongarra discussed his creation and leadership at the University of Tennessee of what became a large research group, the Innovative Computing Laboratory, its sources of funding, and the systems it developed. His next major project was LAPACK, a collaboration with Jim Demmel and several others to create a new library for linear algebra and matrix functions that would maximize performance on shared memory parallel and vector architectures. LAPACK worked closely with Level 2 and Level 3 BLAS, giving a higher level of abstraction from hardware differences and so combining high performance with portability. This led to ScaLAPACK, a version of LAPACK usable on distributed memory systems, and to a more general interest in message passing standards for portable software which Dongarra pursued through the PVM project and the MPI standard. Dongarra also discusses then-current and recent projects, including ATLAS (a system to automatically generate optimized code for particular architectures) and NetSolve, a mechanism to make mathematical software capabilities available as services over a network.




Dongarra, Jack J., Interviewee
Haigh, Thomas, Interviewer


SIAM and U.S. Department of Energy

Place of Publication

Knoxville, TN, USA


79 p.



Copyright Holder

Computer History Museum




Argonne National Laboratory; EISPACK; Software; Applied mathematics; Los Alamos National Laboratory; LINPACK (Software library)

Collection Title

Society for Industrial and Applied Mathematics (SIAM) oral history collection


Gift of SIAM and the US Department of Energy

Lot Number