Installation Guide for ScaLAPACK | Center for Big Data Analytics

Abstract: This working note describes how to install and test version 1.7 of ScaLAPACK. This release of ScaLAPACK extends the functionality of the package by providing new routines for the nonsymmetric eigenproblem, divide and conquer symmetric eigenproblem, and improved routines for generalized symmetric eigenproblem and the reduction to tridiagonal form. It also includes PBLAS v2.0. These two-dimensional distributed memory versions of common LAPACK routines rely on calls to the BLAS for local computation, and calls to the PBLAS for global computations. For portability concerns, communication takes place inside the PBLAS through calls to the BLACS. The design of the testing/timing programs for the ScaLAPACK codes is also discussed.

Topics:
Numerical Linear Algebra

Download: pdf

Citation

Installation Guide for ScaLAPACK (pdf, software)
J. Choi, J. Demmel, I. Dhillon, J. Dongarra, S. Ostrouchov, A. Petitet, K. Stanley, D. Walker, R. Whaley.
University of Tennessee Computer Science Technical Report (UT-CS-95-280) UT-CS-95-280, March 1995.
(updated August 31, 2001 (version 1.7) — also available as LAPACK Working Note 93)

Bibtex:
@techreport{choi1995installati, author = "J. Choi AND J. Demmel AND Inderjit S. Dhillon AND J. Dongarra AND S. Ostrouchov AND A. Petitet AND K. Stanley AND D. Walker AND R. C. Whaley", title = "Installation Guide for ScaLAPACK", institution = "University of Tennessee", number = "UT-CS-95-280", year = "1995", month = "mar", note = "; updated August 31, 2001 (version 1.7) — also available as LAPACK Working Note 93", abstract = "This working note describes how to install and test version 1.7 of ScaLAPACK. This release of ScaLAPACK extends the functionality of the package by providing new routines for the nonsymmetric eigenproblem, divide and conquer symmetric eigenproblem, and improved routines for generalized symmetric eigenproblem and the reduction to tridiagonal form. It also includes PBLAS v2.0. These two-dimensional distributed memory versions of common LAPACK routines rely on calls to the BLAS for local computation, and calls to the PBLAS for global computations. For portability concerns, communication takes place inside the PBLAS through calls to the BLACS. The design of the testing/timing programs for the ScaLAPACK codes is also discussed." }