Allow for non standard cosmologies, and improve cosmological time integration #400
Conversation
We need to track explicitly Omega_R (radiation), w0, and wa for time-variable dark energy models. We also need to write to disk the expansion history, so we can verify that it's correct for unusual cosmologies. We've implemented much of that framework here, but will need to test.
Not sure why this is triggered for cosmology_functions.cpp but not io.cpp.
extra comma deleted
Added a computation of the offset in time at the start of the simulation, and a higher accuracy Runge-Kutta timestep scheme to improve the time integration given the typical scale factor step da.
Removing functionality just to get this to pass
This checks the cosmological simulation grid units. Something similar is in Bruno's original cosmology version.
Removing the dens_number_conv a^3 scaling from chemistry_functions.cpp
Default dev branch does not match original cosmo code behavior. Removing the eta_1 logic from hydro_cuda.cu and then changing the eta_1 default value to 0.001 seems to roughly follow Bruno's cosmo verison.
Changes were made to PPMP that affect reconstruction using the SIMPLE integrator, where "#ifdef CTU" was changed to "#ifndef VL". This affects the cosmological simulation hydro evolution, and behavior similar to Bruno's original code is recovered with this change reversed and SIMPLE integration used.
|
After a lot of effort and help from Bruno, the version 1f05b8d reproduces well the original cosmological hydro sims using the same flags. The salient issue that may need discussion is the last commit that undoes changes to PPMP_cuda.cu, where ifdef CTU had been changed to ifndef VL. The last commit undoes this change, since it was affecting the cosmological hydro integration, and these portions of the PPMP scheme now only apply to CTU again (not VL or SIMPLE). @evaneschneider do you know why this change was made? Is it OK to keep this off whenever SIMPLE is used, or should I add an ifndef COSMOLOGY and revert the changes back to ifndef VL? |
|
@brantr I do think this needs some discussion. Perhaps part of the problem is that I don't really understand what SIMPLE should do. As I've mentioned a couple times, I think SIMPLE is numerically unstable in more than 1D, since I thought it was just the first half of CTU, and removing the characteristic projections and time evolution of the interface states makes it numerically unstable in 1D as well, so I probably thought I was helping when I noticed that step had been eliminated. (I do remember showing some KH tests in the CAAR slack that demonstrated instabilities that went away when the characteristic projection step was turned back on.) That said, if you all have worked out that simple works without those steps, we can (and perhaps should) get rid of all of that code (and I think we should be consistent here, either we have it in all the reconstructions, or we have it in none). I think resolving this question is best done through a conversation, since it seems like maybe you all did some development work in creating simple that I'm unaware of / do not understand. |
|
Let's discuss offline, but then when we resolve we can post back here for prosperity. |
|
When you get a chance @brantr can you re-run clang-format? I was going to try to investigate the linting error but right now there are a bunch of violations in a bunch of files. |
|
@brantr Are you sure you're running the right version of clang-format? There are still a bunch of errors in a number of files, but a lot of them seem to be simple things (like not removing end spaces or not getting equalities in line) that should just be taken care of when you run the formatting, so I don't really understand why it's not working. |
|
BTW, other than the formatting errors, everything seems okay on this branch except the new cosmology system test, which makes sense. Given that I assume this version has been verified for some forms of correctness, I'm happy to merge this and then rerun the system test and swap out the test data. |
|
pre-commit.ci autofix |
for more information, see https://pre-commit.ci
|
Oh wow! it looks like it precomit worked! That's extremely encouraging! |
We want to include non-standard cosmological models that have time-dependent dark energy equation of state. This pull request implements w0,wa parameterization of the DE EOS. It also incorporates a Runge-Kutta integrator to improve the integration of the physical timestep with scale factor, pre-computes the offset in physical time for z<\infty ICs, and writes the expansion history to a file. We also allow for non-zero radiation energy density. I have checked the expansion history agrees to within ~0.1% with independent calculations.