Pyfr partition tavg plugin

Dear PyFR developers,

I would like to partition solutions of the tavg plugin in order to make easier the post-treatment on it.

However when running pyfr partition on the solution of this plugin, I have the following message error:

Traceback (most recent call last):
File “/home/anthony/anaconda3/bin/pyfr”, line 11, in
load_entry_point(‘pyfr==1.10.0’, ‘console_scripts’, ‘pyfr’)()
File “/home/anthony/anaconda3/lib/python3.6/site-packages/pyfr-1.10.0-py3.6.egg/pyfr/main.py”, line 112, in main
File “/home/anthony/anaconda3/lib/python3.6/site-packages/pyfr-1.10.0-py3.6.egg/pyfr/main.py”, line 173, in process_partition
File “/home/anthony/anaconda3/lib/python3.6/site-packages/pyfr-1.10.0-py3.6.egg/pyfr/main.py”, line 166, in
File “/home/anthony/anaconda3/lib/python3.6/site-packages/pyfr-1.10.0-py3.6.egg/pyfr/partitioners/base.py”, line 307, in partition_soln
File “/home/anthony/anaconda3/lib/python3.6/site-packages/pyfr-1.10.0-py3.6.egg/pyfr/partitioners/base.py”, line 110, in _combine_soln_parts
File “/home/anthony/anaconda3/lib/python3.6/site-packages/pyfr-1.10.0-py3.6.egg/pyfr/util.py”, line 35, in call
File “/home/anthony/anaconda3/lib/python3.6/site-packages/pyfr-1.10.0-py3.6.egg/pyfr/readers/native.py”, line 46, in array_info
ValueError: max() arg is an empty sequence

I specify that this operation works with the writer plugin.

So is it because the keys of the tavg and the writer plugin solutions have different names ? Are there basic changes in the code that could enable to partition solutions the tavg plugin ?

Also, I would like to know another thing: imagine that two solutions are issued from initially the same gmsh mesh but that we partitioned the mesh differently. When we run the pyfr partition 1 mesh.pyfrm solution.pyfrs for both solutions do the elements indexes match or are they different ?

Indeed, I would like to perform some post-treatment on all the .pyfrs files of the simulation but do not know how to do it when I do a pyfr restart with a mesh partitioned distinctly due to the computational resources that are available at the moment.

Regards,
Anthony Larroque

Hi Anthony,

It is not (currently) possible to repartition time average files. It would not be too difficult to update the code to support this.

In general if you independently partition a pair of solutions and then collapse them both down to a single partition, the element numbering’s will be different. In order to track element numbers you can use the -r flag to partition which will output the renumbering file.

Regards, Freddie.

Hi Freddie,

Thank you for your answers !

Regards,
Anthony Larroque