The Wayback Machine - https://web.archive.org/web/20070513173030/http://www.sgi.com:80/products/software/scsl.html
SCSL (Scientific Computing Software Library)

SCSL is a powerful software library consisting of a wide variety of industry-standard and proprietary scientific and math functions that have been optimized for use on SGI® computer systems. An important enabling technology, SCSL is now provided at no additional cost to SGI customers with IRIX®, as of IRIX release 6.5.15, February 2002. SCSL may be downloaded from the download software page.

Advantages

  • Provides an extensive selection of scientific and mathematical functions
  • Minimizes code development effort, saving the programmer time and effort
  • Maximizes code performance
  • Provides shared memory parallelism
  • Provides same API for all supported SGI platforms for easy migration of code
  • Supports 'de-facto' BLAS (Basic Linear Algebra Subprograms) and LAPACK (Linear Algebra Package) standard
  • Supports 64-bit integer arguments to facilitate porting from legacy platforms
  • Supports single- and double-precision, real and complex data types
  • Callable from Fortran, C and C++
  • Tested extensively for reliability and accuracy

Description

SCSL is a collection of high-performance routines that provide support for mathematical and numerical techniques used in scientific and technical computing. SCSL encompasses a wide variety of industry-standard and proprietary scientific and math functions including numerical linear algebra, large sparse systems of linear equations, and signal processing.

Included in SCSL are:

  • BLAS
    The BLAS library is a collection of some 400 routines providing important building blocks. These operations include, but are not limited to matrix multiply, rank updates, and dot products. Included are:
    • BLAS1 - Vector-vector operations
    • BLAS2 - Matrix-vector operations
    • BLAS3 - Matrix-matrix operations

  • LAPACK
    LAPACK is a portable library of subroutines for solving the most common dense linear algebra problems: systems of linear equations, linear least squares problems, eigenvalue problems, and singular value problems. LAPACK consists of routines for:
    • Symmetric and Nonsymmetric linear systems of equations
    • Symmetric and Nonsymmetric eigenvector/value
    • Singular Value Decomposition (SVD)
    • Linear Least Squares

  • Signal Processing
    The signal processing library consists of routines that perform mixed-radix fast Fourier transforms (FFTs) as well as linear filtering operations such as convolution and correlation. Supported functions include:
    • One-, two- and three-dimensional mixed radix FFTs
    • Multiple one-dimensional mixed radix FFTs
    • One-, and two-dimensional convolutions
    • Multiple one-dimensional convolutions
    • One-, and two-dimensional correlations
    • Multiple one-dimensional correlations

  • Sparse direct solvers
    • Parallelized direct solvers for sparse systems of linear equations with symmetric non-zero structure
    • High-performance out-of-core solver for sparse linear systems of equations

  • Sparse iterative solvers