wiki:NumpyInstall

Version 34 (modified by rwilson, 8 years ago) (diff)

--

Installation procedures for the imminent numpy ANUGA release

This page describes the procedures used to install ANUGA with the support packages which allow you to run numpy ANUGA on Ubuntu 8.10 (32 or 64 bit) and 32 bit Windows (XP or Vista).

This page will exist until the numpy release is finalised, at which time these procedures will be moved to the ANUGA installation guide.

The procedures below use various support packages such as numpy. The files containing the packages may be downloaded from SourceForge in the numpy_support_software package.

Source package installations require you to unpack a *.tgz file. It doesn't matter where you do this, as you don't need the unpacked directory after a successful installation. You could use /tmp, for instance.

These procedures were written before the numpy trunk became the main branch in the repository. Any paths to numpy code mentioned here will be those in the old system. After numpy code goes into the main branch this document will be amended, but you should check the ANUGA wiki front page for announcements.

Ubuntu 8.10 - 32 bit

Using the Ubuntu Synaptic package manager, install the following packages:

subversion
python-dev
python-profiler
g++
python-psyco
python-numpy

Of course, if you prefer using the console, you could do:

sudo install subversion

and so on.

Next, you need to install NetCDF 4.0.1 from the source package:

tar xzf netcdf.tar.gz
cd netcdf-4.0.1
./configure
make check
sudo make install

Now install ScientificPython 2.9.0:

tar xzf ScientificPython-2.9.0.tar.gz
cd ScientificPython-2.9.0
python setup.py build --numpy
sudo python setup.py install

And finally, install matplotlib through Synaptic:

python-matplotlib

Ubuntu 8.10 - 64 bit

Using the Ubuntu Synaptic package manager, install the following packages:

subversion
python-dev
python-profiler
g++
python-numpy

Of course, if you prefer using the console, you could do:

sudo install subversion

and so on.

You need to specify a special flag for source package builds:

export CFLAGS=-fPIC

Note: you must execute the above console command in any terminal in which you do the NetCDF and ScientificPython source builds (below).

Next, you need to install NetCDF 4.0.1 from the source package:

tar xzf netcdf.tar.gz
cd netcdf-4.0.1
./configure
make check
sudo make install

Now install ScientificPython 2.9.0:

tar xzf ScientificPython-2.9.0.tar.gz
cd ScientificPython-2.9.0
python setup.py build --numpy
sudo python setup.py install

And finally, install matplotlib through Synaptic:

python-matplotlib

Windows

Install python 2.5:

execute python-2.5.4.msi

and add ';C:\python25' to the end of the PATH environmental variable.

Install the MinGW package (requires an internet connection) by:

execute MinGW-5.1.4.exe
 . install MinGW base tools
 . g++ compiler
 . MinGW Make

and add ';C:\MinGW\bin' to the end of the PATH environmental variable.

Install numpy:

execute numpy-1.3.0-win32-superpack-python2.5.exe

install NetCDF4:

execute netCDF_binary_4.0.exe

and add ';C:\netcdf4\bin' to the end of the PATH environmental variable

Install ScientificPython:

execute ScientificPython-2.9.0.win32-py2.5.exe

install psyco:

execute psyco-1.6.win32-py25.exe

This is not strictly required, but it's small and speeds up your ANUGA system.

Install matplotlib:

execute matplotlib-0.98.5.2.win32-py2.5.exe

Testing the installation

Ubuntu

Before testing, you must have installed the ANUGA system. How to get the source code is documented elsewhere, but to be brief, you can either download the latest numpy release package and unpack it into the suggested place (/usr/lib/python2.5/site-packages), or you can get the numpy source tree through subversion.

If you get the numpy ANUGA system through subversion, you will probably download the branches/numpy source tree to a place of your choice, say ~/ANUGA. In this case you need to set the PYTHONPATH environment variable to run ANUGA:

cd ~/ANUGA
svn co https://datamining.anu.edu.au/svn/ga/branches/numpy
svn co https://datamining.anu.edu.au/svn/ga/branches/numpy_anuga_validation  # need to validate ANUGA
export PYTHONPATH=~/ANUGA/numpy

Note that the PYTHONPATH environment variable needs to be persistent, so you might consider putting the above command in the appropriate file (~/.bashrc, etc).

To test your ANUGA install, do:

cd ~/ANUGA/numpy/anuga
python compile_all.py
python test_all.py

To validate ANUGA, do:

cd ~/ANUGA/numpy_anuga_validation/automated_validation_tests
python validate_all.py

Windows

Before testing, you must have installed the ANUGA system. How to get the source code is documented elsewhere, but to be brief, you can either download the latest numpy release package and unpack it into the suggested place (/usr/lib/python2.5/site-packages), or you can get the numpy source tree through subversion.

If you get the numpy ANUGA system through subversion, you will probably download the branches/numpy source tree to a place of your choice, say ~/ANUGA. In this case you need to set the PYTHONPATH environment variable to run ANUGA:

cd ~/ANUGA
svn co https://datamining.anu.edu.au/svn/ga/branches/numpy
svn co https://datamining.anu.edu.au/svn/ga/branches/numpy_anuga_validation  # need to validate ANUGA
export PYTHONPATH=~/ANUGA/numpy

To test your ANUGA install, do:

cd ~/ANUGA/numpy/anuga
python compile_all.py
python test_all.py

To validate ANUGA, do:

cd ~/ANUGA/numpy_anuga_validation/automated_validation_tests
python validate_all.py