Rotation curves are calculated and plotted using the ~pynbody.analysis.profile.Profile object. For a tutorial on profiles see profile.
tutorials/example_code/rotcurve.py
The rotation curve is calculated by calculating the forces in a plane. The force calculation is a direct N2 calculation, so it takes a while and it is therefore done in C. It is even faster if you use the parallel Open-MP version. This was installed automatically if pynbody detected an Open-MP C compiler during setup. To see whether this happened or not, you can ask how many cores your machine has:
In [4]: import pynbody
In [4]: pynbody.openmp.get_cpus()
If this returns 1 but you have more than one core, pynbody has not been installed with OpenMP support (check your compiler
<pynbody-installation-must-haves>, especially on Mac OS - then raise an issue if you can't fix the problem) . If it returns the number of cores on your machine, you're in luck.
Assuming OpenMP support is enabled, the actual number of cores used by pynbody is determined by the configuration option number_of_threads, which is the number of CPUs detected on your machine by default. If you want to reduce this (e.g. you are running on a login node or have multiple analyses going on in parallel), you can specify the number of cores explicitly:
In [5]: pynbody.config['number_of_threads'] = 2
Now all gravity calculations will use the parallel gravity calculation on only 2 cpus. Note that the number of threads you specify in this option will also be the default for other routines, such as pynbody.plot.sph.image. See threads.
If you want to make your configuration changes permanent, create a file called .pynbodyrc in your home directory with the lines
[general]
number_of_threads: 2See configuration for more information.