Back to files

README.SPARSE:

        Rlab2 has a sparse-matrix type. All of the basic functions are
        builtin and ready to run, with the exception of the
        linear-algebra operations. 

        *************

        The SuperLU package:

               http://www.cs.berkeley.edu/~xiaoye/superlu.html

        is a good package for solving sparse systems of equations. The
        package is written in C, and will compile on most platforms
        with an ANSI-C compiler (gcc will do). Once SuperLU is
        installed on your system (preferably in /usr/local/lib, and
        /usr/local/include) you can configure RLaB with the
        `--enable-SUPERLU' option.

        *************

        The UMFPACK routines, or the Harwell Subroutine Library, can
        also supply the function of sparse matrix factorization. Both
        of these libraries carry special licenses, and cannot be
        distributed with Rlab. UMFPACK is available from netlib, and
        information on the Harwell Subroutine Library is available at:

                http://www.cis.rl.ac.uk/struct/ARCD/NUM.html

        Both the HSL and UMFPACK have identical interfaces, and should
        be usable from Rlab2 with little or no modification (I can't
        afford the HSL so I can't be sure).

        *************

        It is possible, fairly easy actually, to create an interface
        from Rlab2 to other sparse matrix linear algebra libraries.

        UMFPACK Notes: If you want to use UMFPACK, you will need to
        compile the the double precision (umd*.f), and complex16
        (umz.f) files. The output files, umd2p?.f and umz2p?.f should
        be converted into "dummy" subroutines by removing all of the
        executable statements. Once you have built libumfpack.a (or
        libumfpack.so), the configure Rlab with the option:
        --enable-UMFPACK