Use a smaller number of neighbours and larger h-tolerance for the stars in the…

Use a smaller number of neighbours and larger h-tolerance for the stars in the EAGLE examples to follow what was done in the Ref model.

48 \pm 1 ngbs for the stars when using the Wendland-C2 kernel. Any thoughts?

added 1 commit

• 7cca29cb - When running in verbose-mode scheduler_splittasks prints out the splitting limits.

Compare with previous version

Can do. I don't really understand why we'd do this; does it really speed things up? Maybe 5% for the stars? ¯\＿(ツ)＿/¯

Ideally I'd like to run with a 'nice' number for eta, like 1.15 and 1.2 for the hydro just to stick it to the idea of requiring an integer number of neighbours.

I don't remember what the rational was. Likely that people were worried that we would do feedback/enrichment too far from a star given the more extended wings of the kernel and the absence of weighting.

I agree with your points. But for now I think we should stick to what we had for comparison purposes until we are confident we can reproduce results.

If there is no weighting then the only acceptable answer that I can see is eta = 1.0. But as you said let's go with this value for now.

Well. You still want to smooth the calculation of other quantities. Like the gas density and other things.

Hmrmmrmmmmm ok. This is not the hill for me to die on :).

assigned to @matthieu

merged

mentioned in commit 6dbaa8c7