[FLASH-USERS] Possibility of restarting FLASH from a selected region

[Seyit Hocuk]

Dear Kevin,

" I am not that familiar with the details of ENZO, but from what I have 
seen I think it is a good code and if it does what you need, give it a 
shot! "

Nor am I, but I will not switch to ENZO after spending so much time on 
FLASH. FLASH is in my opinion a better code for the stuff that I do. :)

Kind Regards,
Seyit


Kevin Olson wrote:
> Dear Seyit,
>
> Since you are using FLASH2.5, what I suggested before will not work 
> since the version of PARAMESH that is used does not allow more than 10 
> refinement levels.  Newer versions PARAMESH allow more refinement 
> levels and are only supported in FLASH 3.x.
>
> I don't think doing what you want to do will help with regards to the 
> timestep problem.
> You would still need to take smaller timesteps in your 'zoomed' in 
> region.
>
> Also, I and others have found the almost all the time computing done 
> in FLASH is done on the finest meshes.  So, derefining the mesh in 
> uninteresting regions of your simulation and keeping them shouldn't be 
> a severe performance hit at all.
>
> I am not that familiar with the details of ENZO, but from what I have 
> seen I think it is a good code and if it does what you need, give it a 
> shot!
>
> Best,
> Kevin Olson
>
> On Mar 12, 2009, at 10:14 AM, Seyit Hocuk wrote:
>
>> Dear Kevin,
>>
>> I am using FLASH 2.5 still. However, 50! levels? That's insane :).
>> Refining is especially interesting (my opinion) if you do 
>> structure/star formation where you have a core collapse and the 
>> density, temperature and pressure increases thereby reducing the 
>> timestep. In effect, for me, I can not go beyond level of ~10 
>> refinement. However, starting from a smaller boxsize (in a specific 
>> region) would make it possible to do the same resolution (say 10) and 
>> result in a detailed image of that area. Of course, refining only 
>> this selected region and adapting the mesh from 10-20, is more ideal 
>> (boundary conditions and stuff), it is much more computer demanding 
>> (even though the number of blocks do not seem increase much) and 
>> impossible to achieve for me.
>>
>> I have the impression that other codes, like ENZO, can do this more 
>> easily.
>>
>> Kind Regards,
>> Seyit
>>
>>
>>
>> Kevin Olson wrote:
>>> Dear Seyit,
>>>
>>> What version of FLASH are you using?  If you are using version 3.0 
>>> or above then you can refine up to 50 levels.
>>>
>>> I don't think FLASH supports specifically what you are asking for.  
>>> You could try just keep the computational domain as it is, derefine 
>>> in the regions where you are not interested and refine in the region 
>>> where you are interested.  An advantage of doing this is that the 
>>> original boundary conditions in your refined regions will be 
>>> preserved.  A disadvantage is that the floating point computations 
>>> may loose accuracy.
>>>
>>> Best,
>>> Kevin Olson
>>>
>>> On Mar 12, 2009, at 9:15 AM, Seyit Hocuk wrote:
>>>
>>>> Hi Community,
>>>>
>>>> I was wondering if it is possible to start a new Flash simulation 
>>>> from an end point of an older simulation by just selecting a part 
>>>> of the data, perhaps by a child block/level. This would now be the 
>>>> top block and act as the new initial conditions, forgetting 
>>>> everything else. I hope I am being clear in this. This way we could 
>>>> go much higher in resolution, without increasing the computational 
>>>> demands, by zooming in on an area and restarting simulation from 
>>>> there.
>>>>
>>>> Thanks,
>>>> Seyit
>>>>
>>>
>>
>