[Mesa-users] D_omega_flag, am_nu_rot_flag, and a possible memory leak

[Evan Bauer]

Hi Avishai,

I’ll leave the rotation physics to somebody with more expertise, but I do have some comments on the possible memory leak. I’m not aware of this being a known issue. A first step that might be useful to us is to run the case that uses 10GB under valgrind to see if it detects any obvious memory leaks. In the work directory that is configured for the 10GB run, you should be able to just run something like this:
	valgrind --leak-check=full --error-limit=no ./star | tee output_valgrind.txt

valgrind is included in the mesasdk, so that should just run out of the box. It will probably take around 100x longer than your MESA run without valgrind. Is that feasible? If so, it should be useful to see the output_valgrind.txt file.

Cheers,
Evan

> On Dec 12, 2018, at 8:56 AM, agilkis--- via Mesa-users <mesa-users at lists.mesastar.org> wrote:
> 
> Hi mesa-users,
> 
> I am looking into core rotation rates of massive stars and the influence of the Spruit-Tayler (ST) dynamo on them. Following a question about rotational mixing during the MESA Summer School, Frank suggested to me using D_omega_flag to get smoother profiles. This seemed to do the job at first, but further runs seemed to raise a couple of issues:
> 1) I noticed that when using D_omega_flag, the core rotation rate becomes faster. I think this is because for the ST dynamo mixing of chemical elements is calculated separately from the angular momentum transport, and using D_omega_flag (with use_D_omega_for_am_nu_rot set to its default value of .true.) results in an underestimation of the core-envelope coupling by magnetic fields (according to Spruit 2002, mixing of elements is less efficient than the angular momentum transport).
> 2) To fix the above, I tried to use am_nu_rot_flag as well as D_omega_flag. This gives a reasonable core rotation rate, similar to the case with no smoothing. However, I noticed that the memory usage in this case kept increasing throughout various runs, reaching up to 10GB in some cases. Without am_nu_rot_flag, the runs used between 1GB and 2GB.
> 
> Attached is a figure of the rotation profile at core helium depletion for no smoothing (inlist0), usingD_omega_flag (inlist1), and using both D_omega_flag and am_nu_rot_flag (inlist2).
> 
> Is the memory leak a known issue? If it is, has it been fixed in a more advanced version of MESA? What would be recommended to do about the smoothing of the profile? I think the current options (version 10398) are (A) to not do smoothing; (B) smooth only with D_omega_flag and set use_D_omega_for_am_nu_rot to .false.
> 
> Thanks!
> Avishai
> <M15V100deplHe4Omega.png><M15V100HR.png><M15V100vrot.png><inlist0><inlist1><inlist2>_______________________________________________
> mesa-users at lists.mesastar.org <mailto:mesa-users at lists.mesastar.org>
> https://lists.mesastar.org/mailman/listinfo/mesa-users <https://lists.mesastar.org/mailman/listinfo/mesa-users>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20181212/5a12f0c7/attachment.html>