Installing GPAW (Python DFT package) on a computing cluster

Question

I'm in the middle of installing GPAW on a cluster but I'm running into problems building the "gpaw-python" interpreter, its job is to allow users to execute parallelised GPAW calculations. Following the Developer Installation guide here, I've downloaded the most recent stable package and got it working fully on my local desktop, including the gpaw-python interpreter, and currently have the gpaw.tar file unzipped and ready to go on the cluster, which I am accessing through ssh.

The specific part where the problem occurs is building the C extensions, as seen at the start of the developer guide. I've ensured that I have mpicc installed as required and it links to the command line ok, albeit this is through a software module on the cluster which is available to all users (this version is gcc/4.6.3/openmpi/64/1.6.1). Note the caveat in the developer installation guide requiring mpicc in order to construct the gpaw-python interpreter.

When I run

python setup.py build_ext 2>&1 | tee build_ext.log

Everything looks okay and the expected location of the "gpaw-python" executable is built in

   $GPAW_HOME/build/lib.<platform>-<python_version>/

but doesn't anything, so I would guess the installation is missing mpicc somehow. When I tried installing GPAW locally there wasn't a problem doing this so maybe I'm missing some subtlety when working on the cluster?

Answer 1

Problem can be fixed by adding this around line 395 in the config.py file, inside the build_interpreter function:

define_macros.append(('_GNU_SOURCE', '1'))

where the other define_macros.append commands reside.