You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
SOAP could be sped up by feeding pre-sorted particles to the property calculations.
Currently the function process_single_halo() in halo_tasks.py finds all of the particles in some radius around the halo centre and feeds them to the property calculations from SO_properties.py, aperture_properties.py and subhalo_properties.py. The radius used is the largest radius needed by any of the property calculations. All property calculations receive the same set of particles, and all of the property calculations start by sorting the particles by radius.
There are two opportunities to speed up the code here:
The sorting by radius could be done once before any of the property calculations are called
Each property calculation could be passed just the particles it needs. Smaller apertures would then process many fewer particles.
To do this we would probably need to factor out the *ParticleData classes into a single class to be set up before calling any of the property calculations.
The text was updated successfully, but these errors were encountered:
SOAP could be sped up by feeding pre-sorted particles to the property calculations.
Currently the function process_single_halo() in halo_tasks.py finds all of the particles in some radius around the halo centre and feeds them to the property calculations from SO_properties.py, aperture_properties.py and subhalo_properties.py. The radius used is the largest radius needed by any of the property calculations. All property calculations receive the same set of particles, and all of the property calculations start by sorting the particles by radius.
There are two opportunities to speed up the code here:
To do this we would probably need to factor out the *ParticleData classes into a single class to be set up before calling any of the property calculations.
The text was updated successfully, but these errors were encountered: