Installation

Installation
============

Required software:
1. any C/C++ and MPI
2. GSL (GNU Scientific Library, version 1.11) - for benchmarks (statistics),
   for numerical partitioners
3. any BLAS (Basic Linear Algebra Subprograms) - for routines (default: GSL CBLAS)

Optional software:
1. NLopt (a library for nonlinear optimization) -
   for numerical partitioners
2. Gnuplot (An Interactive Plotting Program) - for performance diagrams
3. CUDA Toolkit 4.0/4.1 - for CUDA BLAS
4. any Python - for CUDA + autotools integration

MPI:
Should be configured with support of C++ and shared libraries

NLopt:
Download the latest version from http://ab-initio.mit.edu/wiki/index.php/NLopt
Configure with C++
$ ./configure --with-cxx

For users
---------

Download the latest version from http://hcl.ucd.ie/project/fupermod

$ tar -zxvf fupermod-X.X.X.tar.gz
$ cd fupermod-X.X.X
$ ./configure
$ make all install

See documentation Quick Start Guide.

Configuration
-------------

Packages:
  --with-gsl-dir=DIR      GNU Scientific Library directory
  --with-nlopt-dir=DIR    NLopt directory
  --with-blas=NAME        BLAS (Basic Linear Algebra Subprograms): gsl, atlas,
                          goto, mkl, acml, cuda (default: gsl)
  --with-atlas-dir=DIR    Automatically Tuned Linear Algebra Software
                          directory
  --with-goto-dir=DIR     Goto BLAS directory
  --with-mkl-dir=DIR      Math Kernel Library directory
  --with-acml-dir=DIR     AMD Core Math Library directory
  --with-cuda-dir=DIR     CUDA directory

Check configure options:
$ ./configure -h

For developers
--------------

Required software:
1. Subversion
2. GNU autotools
3. Doxygen, Graphviz and any TeX - optional (for reference manual)

$ svn co svn+ssh://gforge.ucd.ie/var/lib/gforge/chroot/scmrepos/svn/fupermod/trunk fupermod
$ cd fupermod
$ autoreconf -i
$ ./configure --enable-debug
$ make all

To create a package:
$ svn log -v > ChangeLog
$ make dist