Smoothed metallicity

Smoothed Metallicity is implemented for Gear. I still need to ensure that the code compile with none and EAGLE, but will do it once you accept the modifications.

You have a new example in SmoothedMetallicity that uses a glass with two different metallicities with some gaussian noise. At the discontinuity (+- the smoothing length), the transition is smoothed and gives the following graph

In this case, I am plotting in red the particles and in blue the theoretical solution (+- 1 sigma of the noise). The three plots corresponds to 3 different elements (I do not want to follow them all as it requires 9 graphs)

@aborissov, the code should be compatible with EAGLE, therefore you will just need to copy/paste

changed milestone to %Chemistry Implementation

changed the description

added 1 commit

• 4c4631a3 - Fix comment

Compare with previous version

Looks good to me in general. Great stuff!

Will have a more detailed look over the weekend. One thing is that it needs to compile at least with chemistry==none (no need for eagle fixes at this stage) otherwise the automated tests will scream.

One more thing you could do to check your implementation is to add the call to the chemistry_iact functions to tests/testSymmetry.c. This unit test explicitely checks that the non-symmetric and symmetric versions are compatible.

For the test case, you can run it for 1 step using "-n 1" in the command line. That should enough for the smoothed metallicites to be computed. You can then use the snapshot number 1 to test it.

added 2 commits

• 2e709b43 - Add chemistry in symmetry test
• dc32a10e - Do only 1 step in SmoothedMetallicity

Compare with previous version

added 1 commit

• d62733f6 - Update doc string for gear and add missing functions to chemistry/none

Compare with previous version

I have:

• updated and tested the chemistry==none
• add chemistry in the symmetric test (but I am not initializing the chemistry data). Currently we do not need the chemistry data, therefore I did not try to initialize it, but I can do it if you wish.
• updated SmoothedMetallicity with -n 1
• updated the wiki hydro test page with the smoothed metallicity

I forgot to mention that I did not implemented the chemistry with Gadget sph and vectorization as I understood that James will remove it.

added 1 commit

• 1a5338fa - Formating

Compare with previous version

unmarked as a Work In Progress

added 1 commit

• 2613add6 - Add eagle iact

Compare with previous version

Thanks!

One quick question about your test. What am I supposed to see on the different panels? Would it be possible to plot the metallicity vs. the smoothed metallicity to show that it is indeed smooth?

Also, do you think we will actually need to pass the full chemistry_data structure to the interaction itself?

added 19 commits

• 2613add6...3cc0dba0 - 15 commits from branch master
• b2d96158 - Merge branch 'master' into smooth_metal
• 270acf0c - Make the run scripts executable.
• a7239fd0 - No need to rescale the smoothing lengths from the glass file.
• 286a53ff - Better way of including the chemistry data in the runner.

Compare with previous version

added 1 commit

• 8e7f248f - Added empty interaction functions for the EAGLE chemistry model.

Compare with previous version

I have done a small set of changes and added empty functions for the EAGLE model to compile. Let me know if you disagree with anything.

added 1 commit

• 4afbdf04 - Removed unnecessary includes. Documentation fixes.

Compare with previous version

added 1 commit

• 996ecde4 - Update SmoothedMetallicity plots

Compare with previous version

I am fine with your modifications.

I have updated the graphs in order to plot what you required. The idea of my old plot was to show that the dispersion is smaller with smoothed metallicity than the input (in blue) and to show the linear transition between the two parts.

Here is an updated plot: For the first one, the black line is the relation x=y and each color corresponds to a different element.

For the interaction, I am not sure if we need it and I do not really see a case where it could be useful. I think if we wish to use them, it will be done in a new task. Maybe we can wait a little bit on the development of the chemistry before removing it?

Excellent, thanks! I think that first plot shows exactly that it works. And for the original plot I was wondering what the difference was between the three panels. It's all clear now!

Let me think about the extra parameter to the interaction functions. Passing in extra things, especially with the pointer de-referenciation here can lead to lower levels of optimisation.