Special installation instructions for various platforms
=======================================================

Win32 using MS VC++ 6.0 and Intel MKL
-------------------------------------

PySparse was successfully built with Microsoft Visual C++ 6.0 together
with the Intel Math Kernel Library (MKL).

When adjusting "setup.py" you can use the settings under "elif hostname
== 'Rivendell':" and change the library_dirs_list according to your
MKL installation.

Before running the setup script please make sure that the VC++
executables are in your PATH environment variable.

It is also important that you include the directory containing the MKL
DLLs in your PATH environment variable. Otherwise runtime errors will
occur when PySparse is imported and called.


Win32 using MinGW
-----------------

1. Follow the instructions at http://sebsauvage.net/python/mingw.html
   to

   - install MinGW

   - build libpythonXX.a from python.dll

   - Note that enthought python comes with libpythonXX.a

2. Install BLAS/LAPACK

   - Download the LAPACK and BLAS sources from
     http://www.netlib.org.

   - Follow the included instructions to build the libraries using the
     MinGW development tools.

   - Install the libraries in your MinGW tree.

   - Pre-compiled libraries for windows are available in Tools/win32_blas_lapack

3. Build pysparse

   - set PATH=C:\mingw\bin

   - Customize setup.py (this should not now be necessary):

     Your should use the following settings:
	libraries_list = ['lapack', 'blas', 'g2c']
    	superlu_defs = [("NO_TIMER",1), ('USE_VENDOR_BLAS',1)]

   - Build build and install:

     C:\pythonXX\python setup.py build --compiler mingw32 bdist --format=wininst

   - Build a distribution for windows:

     C:\pythonXX\python setup.py build --compiler=mingw32 --install

