Python Interface¶
Wrapper for the C Interface diverge.h, generated with some help of
ctypesgen. It is heavily based on wrapping C arrays as numpy structured arrays.
Therefore, some documentation for these (and how to use them from the
diverge python module) is included in the section Structured Arrays.
The documentation only includes those functions that exist on top of the C Interface. Everything else is documented there. For usage examples, see the python examples shipped with diverge.tar.gz or found in the examples/ directory of the repo . They generically work similar to the following skeleton:
import diverge
diverge.init()
model = diverge.model_init()
model.contents.MEMBER = VALUE
# and so on
diverge.model_internals_common( model )
diverge.model_internals_tu( model, 2.01 )
step = diverge.flow_step_init_any( model, "PCD" )
# some flow loop and do whatever you like with step/model
diverge.flow_step_free( step )
diverge.model_free( model )
diverge.finalize()
Notice how the diverge_ prefix disappeared into the module name. This holds
for all functions that hold this prefix. Others keep their name.
Note
Installation and Extension
On GNU/Linux, the python library will search for libdivERGe.so in your
LD_LIBRARY_PATH with precedence over the shipped version (located
somewhere in your PYTHONPATH). This makes it easy to swap out the backend by
pointing LD_LIBRARY_PATH to a location with the desired libdivERGe.so:
LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/gpu/version/ python script.py
Note
Python Developers
Installing the library with pip install -e is recommended for not
having to reinstall the package after every change. On some Linux
systems, you may have to fiddle around with pip’s flags (e.g. pip
install --editable . --config-settings editable_mode=strict
--break-system-packages).