Software SPARSEKIT

The library software SparseKIT, which has been developed by Youcef Saad (last update 2004), has main page here, currently to the first write of this document (Out 2011).
This library of fortran 90 subroutines is able to perform a small variety of Matrix Linear Algebra operations with sparse matrices in various compressed formats, among them the well-known Harwell-Boeing Exchange Format. This is of my direct research interest, in order to make sparse computation on the test matrices of the Harwell-Boeing Sets of the Matrix Market web repository (link here).

In the website above, you can find instructions on how to install sparsekit (as a library to be linked through -lsparsekit compilation directive) in your computer. Particularly, sparsekit installation requires f90split, which can be found here , together with installation notes. Since installation does not go as smooth as you probably want, if you need further help, please take a look at my own notes further below in this document.

Installing SparseKIT in your Linux computer

Before you begin, be sure that you are running as root, and that you have gfortran installed.
  1. Download the files f90split.f90 and sparsekit.f90 from the websites above
  2. Run mkdir /usr/local/sparsekit to create permanent directory. Place the two .f90 files there.
  3. Run cd /usr/local/sparsekit
  4. Run gfortran -g f90split.f90 -o f90split (you should notice the creation of an executable file f90split using ls)
  5. Optionally, run ln -s /usr/local/sparsekit/f90split /usr/bin/f90split to create a symbolic link that allows you to call your executable f90split from anywhere in your computer
  6. Run mkdir src; cp sparsekit.f90 src; cd src
  7. Run ../f90split sparsekit.f90 (you will see a report, and at the end, statement on Normal end of execution)
  8. Run for FILE in `ls -1 *.f90`; do gfortran -c -g $FILE; done
  9. Run cd ..; ar qc libsparsekit.a src/*.o (should notice the creation of file libsparsekit.a)
  10. Run ln -s /usr/local/sparsekit/libsparsekit.a /usr/lib/libsparsekit.a to place symbolic link to your just-created library
  11. Optionally, you can run rm -rf src to remove unnecessary files
JBC, Out 2011