Change log

Log of major changes to the code.

28-Apr-24

v0.4 Altered the population of the output buffers for halos, subhalos and galaxies to be done using a range, rather than looping over objects one at a time. This reduces the time taken to an insignificant amount. Testing on the first 10 and 1000 halos in Mill/pygal_063_5.hdf5 gives identical results.

18-Apr-24

v0.3 Halo and subhalo class instances have been removed and converted to numpy arrays. driver.py has been split into push_snap.py and process_snap.c. This gives a speedup of more than a factor of 2. Testing on the first 10 and 1000 halos in Mill/pygal_063_5.hdf5 gives identical results.

20-Feb-24

v0.2 All the galaxy astrophysics routines have been converted to C. This provides a significant speed-up compared to version v0.1. Testing on the first 10 halos in Mill/pygal_063_5.hdf5 gives identical results.

8-Dec-23

v0.1 Froze the python version of the code with the tag v0.1. This is essentially the new_input_format version as described below, but with some optimisation.

28-Nov-23

Have t​​wo working versions in separate branches:

  • main – works with existing MEGA input. That is deficient in many ways and the input format needs to be changed. This has been done in…

  • new_input_format – required entries in the graph input file are much better defined with cleaner names. A helper function has been written to convert Millennium output tree files into that format (but without the ability to trace orphaned halos).

The basic code contains the following astrophysics:

  • Pushing of material from one snapshot to the next: halos, subhalos, galaxies.

  • Accretion of baryons onto halos.

  • Reaccretion of ejected gas.

  • Cooling of gas from halos onto subhalos and subhalos onto galaxies.

  • Merging of galaxies within subhalos, with possible starburst.

  • Star formation and feedback.

  • AGN feedback.

The code currently has the following complexity:

  • Star formation histories.

  • Not resolved discs.

  • Not complex metallicities, just a single overall metallicity.

Preliminary routines have been written to analyse Halo, subhalo and galaxy output. These show that the code performs broadly as expected but with some outliers: I suspect that this is a problem with the input graphs/trees from Millennium.

Routines have been implemented to monitor cpu memory usage (see separate document). It is fair to say that both of these seem very poor at the moment compared to a pure C implementation and the next step is to improve this to demonstrate that the code will have a practical use.