Welcome to the NLPy optimization toolkit for Python. If you read documentation, the next section is for you. Otherwise, skip to the Installation section below.
The NLPy documentation is typeset using the Sphinx documentation system `http://sphinx.pocoo.org`_ which is particularly suited to documenting Python packages.
The manual in PDF format is in the doc subfolder. The HTML documentation is accessible by pointing your browser to doc/build/contents.html.
Check the paper NLPy---A Large-Scale Optimization Toolkit in Python by Dominique Orban available from `http://www.gerad.ca`_.
If the user so desires, the documentation may be generated afresh as follows:
To re-generate the HTML documentation, change to the doc subfolder and type:
make html
To re-generate the PDF documentation, change to the doc subfolder and type:
make latex
Then change to the build/latex subfolder and type:
make all-pdf
For now, just:
cp site.template.cfg site.cfg # Edit site.cfg to adjust to your local settings. python setup.py build python setup.py install [--prefix=...]
To select another C compiler:
python setup.py config --compiler=<name> build
To select another Fortran compiler:
python setup.py config_fc --fcompiler=<name> build
To see a list of available Fortran compilers and their names:
python setup.py config_fc --help-fcompiler
To select both a C and a Fortran compiler:
python setup.py config --compiler=<name> config_fc --fcompiler=<name> build
For example, you can force compilation with gfortran by specifying:
--fcompiler=gnu95
and wit g95 by specifiying:
--fcompiler=g95
(note the subtle name difference.)
On an OpenSuSE system I obtain the following error message with both gfortran and g95:
ld: build/temp.linux-x86_64-2.6/libnlpy_ma27.a(ma27ad.o): relocation R_X86_64_32S against 'a local symbol' can not be used when making a shared object; recompile with -fPIC build/temp.linux-x86_64-2.6/libnlpy_ma27.a: could not read symbols: Bad value
The source of the problem is that on this platform, distutils appears to be compiling a static library without the -fPIC flags, which generates relocatable objects. When subsequently linking against this library to build a shared object, the linker complains. This problem can be resolved by restarting the installation and adding the following flags to the python setup.py command line:
--f77flags="-ffixed-form -fno-second-underscore -fPIC -O" --f90flags="-fno-second-underscore -fPIC -O"