Implement star skipping

added GEAR performance labels

Can you confirm that I am using the correct times here please? I have some doubts about ti_begin.

This function is called in the task timestep after the update of the time_bin.

added 3 commits

• 72db4852 - Update time
• 0d20bdb4 - Update API
• fe1bc4de - Format

Compare with previous version

marked as a Work In Progress

unmarked as a Work In Progress

assigned to @matthieu and unassigned @lhausammann

That's a lot more complicated than I imagined. Let me check it later.

Have you verified that it does not change the physics?

Yes, everything seems fine.

The problem is that we have 2 different time for the supernovae (Ia and II) and that our model is entirely based on the mass. Therefore the logic is a bit longer than expected but not so complicated.

Can you not use a user-defined age above which you downsample?

I am not sure to understand what you mean. In the parameter file, you provide the time limits and then you compute a better estimate?

I do not see the point.

closed

reopened

If age > X then run feedback every N star step. X and N are user provided values.

This is a different problem. Here I am actually computing when I can produce a supernovae.

This is our graph for the supernovae rate. As you can see below 5Myr, between 30Myr-400Myr and after 10-14Gyr (a bit less useful this one), we cannot produce any supernovae: rates.pdf

Currently, we do not wish to downsample. We are looking at smaller system, therefore (I guess that) the downsampling will impact more our dwarfs as there is not enough stars to have a continuous explosion of supernovae.

Ok, so given the plot, there isn't much we can save here anyway since your stars may do something for most of their lifetime (if you plot the x-axis linearly on your plot) and you want this done every (gravity) step potentially.

I agree, but it is already a small win as this function is quickly computed in comparison to the density loop. By the way, I can still improve this by skipping the feedback loop. I will quickly push an update (no need to run an additional test, it is really safe).

added 1 commit

• f36edce0 - Turn off feedback when no SN produced

Compare with previous version

Ok it is done.

When I do not need to explode a supernovae, I am now turning off the feedback in the stellar evolution. This can happen when the time step is sufficiently small and the rate too (I would say that it happens more than 50% of the time in our simulation).

By the way, do you need the stellar evolution after the density loop? It might be more efficient to do it in the kick2 in order to know exactly the stars that will explode.

yes, because we need the density and mass in the star's kernel.