[FLASH-USERS] StirTurb Restart

[Christoph Federrath]

Hi Evan, Hi Klaus,

I'm not sure what exactly the problem is, but I figured that all out long time ago when I switched to FLASH3 and now FLASH4, as I've been running turbulent stirring for years now in FLASH. First, I have to say that the existing "stirturb" module does not produce a purely solenoidal driving field contrary to what the manual says; there is a bug in how the projection in Fourier space is done. I also had to modify the IO routines if I remember correctly to get the restarting fixed. Basically, the modes and other variables were not written to the checkpoint file, and so the simulation did not continue correctly on restart (meaning it probably did restart, but gave results that depended on whether a restart was done or not). I have fixed this, but I went further and modified the stirring module to now read in a predefined forcing sequence, which we used in a code comparison.

My plan was to make the modified turbulence driving module available anyway at some point. If anybody is interested, I can send it to you; works well on restart; supports purely solenoidal or compressive driving and any combination; and allows to define stirring with a predefined sequence (you can always reconstruct the driving sequence afterwards, extend it, etc).

Kind regards,

Christoph

PS: Here are some works using the modified stirring module in FLASH:
http://adsabs.harvard.edu/abs/2010A%26A...512A..81F
http://adsabs.harvard.edu/abs/2010MNRAS.406.1659P
http://adsabs.harvard.edu/abs/2013arXiv1306.3989F

________________________________
Dr. Christoph Federrath
School of Mathematical Sciences,
Monash University,
Clayton, Vic 3800, Australia
+61 3 9905 9760
http://www.ita.uni-heidelberg.de/~chfeder/index.shtml?lang=en


Am 10.09.2013 um 04:54 schrieb Klaus Weide:

> On Mon, 9 Sep 2013, Evan Scannapieco wrote:
> 
>> Hi Folks,
>> 
>> So I have a problem.  I can't do a restart for the StirTurb problem.  
>> I've run StirTurb problems of many kinds for quite a while now, but have 
>> never been able to restart them, neither on our local cluster, nor on 
>> XSEDE resources. This time we must restart, as there is no workaround 
>> for the problem we are trying to do.
> 
> Evan,
> 
> What do you mean with "I can't do a restart"? is there a failure at
> setup time, during compilation, or at runtime, or do you just get the
> wrong results?
> 
> What are the error messages (standard output and log file if you get that 
> far)?
> 
>> Clearly it has something to do with IO_writeUserArray.F90 and 
>> IO_readUserArray.F90 and how they link with the hdf5 and such. Dean 
>> Townsley and Bob Fisher suggested that this might have to with hdf5 
>> flags choice I have tried with and without parallel IO, without and 
>> without the --with-default-api-version=v16 flag on, on both FLASH 3.2 
>> and FLASH 4, And thus far I failed to come up with any combination that 
>> gives successful results.
>> 
>> So I'm wondering if you any one has an idea what is the combination of: 
>> hdf5 version and compiler flags used when building it (Id prefer serial version if possible)
>> MPI version (and type such as openmp, mvapich, mvapich2, etc.)
>> Compiler (hopefully intel is okay) and if so which version of intel compilers? 13?
>> FLASH version and flags (I'd prefer if I can do it with FLASH4, but can live with 3.2 also).
> 
> It all should work unless there is some library mismatch, as long as you 
> are using parallel I/O ('+parallelIO').  Yes, that means you also need
> a parallel build of your HDF5 library.
> 
> 
> Can you run any other Flash test simulation with parallel IO on THIS 
> system?
> 
> Can you run a parallel StirTurb restart (maybe a smaller test) on another 
> system?
> 
> 
> Klaus