Cooling & Energy (U) Based Schemes
It appears that there is some issue with cooling (of any kind) and the energy based schemes. This first appeared when trying to run the SmallCosmoVolume
test with the ANARCHY-PU scheme with EAGLE cooling, but it seems to be more widespread than that. EAGLE cooling gives crazy results (see the below plots; not all of these completed because I killed the jobs or they crashed) with the minimal
, anarchy-pu
, and pressure-energy
schemes.
Minimal & EAGLE
ANARCHY-PU (no diffusion) & EAGLE
ANARCHY-PU (fixed artificial viscosity & no diffusion) & EAGLE
Minimal & Const-Lambda
The modules used on COSMA for these runs are:
1) gnu_comp/7.3.0(default) 4) gsl/2.4(default) 7) python/3.6.5
2) openmpi/3.0.1 5) fftw/3.3.7(default)
3) parallel_hdf5/1.8.20 6) metis/5.1.0-64bit
I used COSMA-6 for these runs, and ran on a single node with either 8 or 16 threads. SWIFT was invoked with ../swift --cosmology --hydro --self-gravity --cooling --temperature --pin --threads=16 small_cosmo_volume.yml
They were performed with git revision 5c774ccc.
The parameter file used for these runs is attached here: small_cosmo_volume.yml
Other Issues
- At the moment, the ANARCHY-PU scheme fails within the first few steps (requesting too low of a time-step) if the energy diffusion is turned on. This could be a cosmology issue with the energy diffusion, or related to this cooling bug.
- The
const-lambda
cooling segfaults ifcooling_tstep_mult: 1.0
is set to anything.