STARSS Star Formation and Feedback #198
Conversation
- seems to work if deposition doesn't leak into ghost zones (need to add refresh)
…deposition still slightly wrong for particles that explode right on the edge of a block.
…y by mass/density for conservative fields
…I think) Long-term evolution still looks slightly asymmetric though Code currently hangs up if feedback isn't the last method in method:list (with refresh enabled).
Ideal test not conserving momentum across block boundaries (might not be feedback related) Other than that, seems to be fully working in cosmo sims
…derdeposition of mass and metals Added tool for generating .block_list file with new_output
…unding cells Changed to be consistent with Enzo version, which also made this change recently
…bug where I could only call feedback as the last method. Storing MMW as double instead of int....................................................................
…gy/potential back to saving particle masses as mass Added debug statement
…ack to saving masses as mass
Moved initialization of deposit fields as `tiny_number` to compute_() to fix bug where if multiple supernovae/stellar wind events occur near a block boundary in a cycle, ghost zone values would get overwritten each time an event occurs for that cycle instead of just adding to a running total of energy, momentum, etc. to deposit.
…kSTARSS moved compute_done() back to compute(). Cosmology simulations are hanging when it's in p_method_feedback_starss_end() for some reason
…sterious "density deposition" bug
|
Hi @WillHicks96 -- the isolated SN test is working great. I have scaled it up to 512^3 and 256 proc on Pleiades with no problems. The cosmology test also starts and has run to z= 8. However, when I try to open the outputs I get a mysterious yt error. I so am uncertain whether this relates to the code of the PR itself or whether its independent. If you recommend that this is separate, I'm ready to approve the PR. Thanks. In [1]: import yt In [2]: ds = yt.load('Dir_COSMO_SF_FB_0100/Dir_COSMO_SF_FB_0100.block_list') In [3]: p = yt.ProjectionPlot(ds, 'x', 'density')
|
|
Thanks, @stefanarridge! I think I'm almost there. There is still some more to do over the next few days before this is ready to merge, mainly finishing the documentation and some more general cleaning. |
|
@tumlinson Thanks for trying out the tests! I've seen that yt error before, and it seems to be fixed in the latest dev version of yt. So if you update your yt, you should hopefully be able to access the data. I notice you're using a version of yt installed with conda. The error happens when yt is trying to create the EDIT: Looks like the issue actually wasn't fixed in yt, and that I just commented out the relevant line in yt when I was frantically preparing for the Enzo-E workshop and then immediately forgot about it. I'll post an issue on the yt github. From what I can tell though, this is a yt issue, and not specifically an issue with this PR |
|
I am on yt 4.1.dev0 and still getting the error, but since this is a yt thing and the Enzo-E tests run I can go ahead and approve. |
No longer need to specify them in a parameter file
|
The code is failing to compile now. |
Removed duplicated code in SolveHydroEquations
|
@jwise77 Think I found the issue. |
…stricted_sn=false, it will only set off a Type IA supernova if it's been determined that no Type II supernovae will go off. How it was written before, no Type IA supernovae would go off if unrestricted_sn=false, which is not the intended behavior.
…nergy formalism. Mainly because it hasn't been tested without dual energy formalism at all
|
@WillHicks96 this looks ready to merge; please do so if there isn't anything left to edit! |
STARSS = Scale-intelligent Terminal-momentum Algorithm for Realistic Stellar Sources
The STARSS algorithm is being actively developed by Azton Wells. Here's a link to his open PR in enzo-dev that has some more info -> enzo-project/enzo-dev#145
This PR adds two new methods
EnzoMethodStarMakerSTARSSandEnzoMethodFeedbackSTARSS, along with some changes and bugfixes to theEnzoMethodStarMakerclass. This depends on PR #143.This is the final form of the
EnzoMethodFire2<StarMaker/Feedback>methods that Andrew Emerick began working on, and I was able to reuse some of his code. There are still some miscellaneous TODO`s in the code (e.g. giving parameters better names and adding more tests), but I think that this is ready for review.I've so far added three files to the input directory:
input/STARSS/SF_FB.incl- This outlines all of the relevant parameters, in the configuration that I've found works best in the cosmology tests I've done so farinput/STARSS/method_feedback_starss.in- This is an ideal test that sets off a single supernova on the corner of a block. This is meant to test that the supernova mass/energy/momentum remains consistent with expectation when the initial deposition crosses block boundariesinput/test_cosmo-dd-SF_FB.in- This is an example that shows how you would call SF+FB in a cosmology simulation. This is a pretty small test, so it won't form its first star particle until around redshift 7.5ish (~cycle 270) with the current parameters. This will take 15-20 minutes on 8 processors. A cosmology test that forms stars more quickly would probably be necessary at some point -- maybe we could have one that restarts from a checkpoint at a redshift where star formation is about to happen.Most of my testing has been within the context of cosmology simulations. Here is a snapshot from a recent 256^3 simulation that shows SF+FB doing things: