Table of Contents
h5py
Installation guide
Running
pip install h5py
gives the error message
h5py/defs.c:8:22: fatal error: pyconfig.h: No such file or directory
this was due to the python development
package not being installed.
Subsequently running
pip install h5py
throws the error message
/usr/include/H5public.h:57:20: fatal error: mpi.h: No such file or directory
this was due to pip
not finding the mpi package. It can be resolved by specifying the CFLAGS
shell variable.
Eventually running
env CFLAGS=-I/usr/lib/openmpi/include pip install h5py
did the trick.
An alternative could be
pip install --global-option "env CFLAGS=-I/usr/lib/openmpi/include" h5py
The alternative did not work for me (Manuel, python2, @ww8stud6) — Manuel Leimberger 2014/08/28 11:19
Installing H5py Parallel HDF5 (Recommended)
Required is a HDF5 library compiled with MPI. Ask your admin or compile it yourself. If that's the case, execute this is in your bash:
export CC=mpicc export HDF5_MPI="ON" pip install h5py --no-binary=h5py h5py
If it has finished without any (major) complaints, test it with this script:
from mpi4py import MPI import h5py rank = MPI.COMM_WORLD.Get_rank() N = MPI.COMM_WORLD.Get_size() f = h5py.File('parallel.hdf5', 'w', driver='mpio', comm=MPI.COMM_WORLD) dset = f.create_dataset('test', (N,), dtype='i') dset[rank] = rank f.close()
by executing: </code> If it has finished without any (major) complaints, test it with this script:
mpiexec -n <number of processors on your computer> python test-h5py-parallel.py
You should have now a hdf5 file of the name “parallel.hdf5” with one dataset inside with the number of processors in c-indexing in sequential order. So for four processors the entries are 0,1,2,3.