[FLASH-USERS] QuickFlash Interpolation Question

[John ZuHone]

Sam,

Visit's manuals are frightfully outdated (at least the ones online), but I think that you are still correct. One can always do a resample using visit but to regrid something uniformly to be put into something else is another matter (which might be still possible but more difficult).

As Nathan was talking about (at least I think this is what he's talking about), in QuickFlash there are some examples of sub-sampling to a uniform grid, particularly in examples/map_uniform, which samples to a  2D uniform mesh. This of course has the associated problems he talked about with the refinement boundaries, but should work fine for some things where this is not a huge issue.

I have taken this routine and expanded it somewhat to include sampling to a 3D mesh, which I can provide for you if this is something that may be useful. 

Best,

John Z

On Feb 17, 2010, at 1:53 PM, Samuel Friedman wrote:

> Dear Tomek,
> 
> I just looked at VisIt and it does not appear to satisfy my needs.  In 
> short, I need to generate a uniform grid of the data.  If you know how to 
> do this with VisIt, I'll listen; I looked through VisIt's manuals and 
> there is no mention about it in the manuals. 
> 
> Sam
> 
> 
> On Wed, 17 Feb 2010, Tomasz Plewa wrote:
> 
>> Sam -
>> 
>> Have you tried VisIt? I am not sure what are you exact needs, but VisIt is
>> more than just visualization tool, but offers a strong and growing capability
>> for data analysis.
>> 
>> VisIt is available at
>> 
>> https://wci.llnl.gov/codes/visit/
>> 
>> and there exists large and active group of users/developers,
>> 
>> http://www.visitusers.org/index.php?title=Main_Page
>> 
>> Tomek
>> --
>> Samuel Friedman wrote:
>>> Dear Nathan,
>>> 
>>> I played around with the code and looked at fidlr's merge_amr routine.  A
>>> few things I noticed:
>>> 
>>> 1)  I managed to create very easily a 3D uniform grid from AMR data by
>>> simply modifying the example routine in map_uniform.  Quite easy.
>>> 2)  fidlr's merge_amr routine calls either the IDL function congrid (in 2D)
>>> or rebin (in 3D).  In both cases, they simply use nearest neighbor sampling.
>>> So I think the "interpolation" should be really easy to implement for
>>> QuickFlash.
>>> 3)  This is harder, but I tried using map_uniform on a simulation that used
>>> a uniform grid and QuickFlash gave me an error message (about an invalid
>>> neighbor code and that mesh information was not available).
>>> 
>>> The main reason I looked into QuickFlash is that it was pretty much the ONLY
>>> set of tools outside of IDL avaiable for analysis.  Ultimately, I am looking
>>> for something without IDL's licensing agreement and fidlr3.0 does not seem
>>> to work properly with GDL.
>>> 
>>> Sam
>>> 
>>> On Wed, 17 Feb 2010, Nathan Hearn wrote:
>>> 
>>> 
>>>> Hi Sam,
>>>> 
>>>>    Thank you very much for your interest in QuickFlash -- providing
>>>> basic tools for developing your own analysis software is what I
>>>> intended by releasing it!  Indeed, the now-quite-old 1.0.0 release has
>>>> only very simple sampling methods for automated extraction of data
>>>> over large areas.  [Massimo: I do apologize for missing your question
>>>> about slicing in October.  I hope that the current discussion is of
>>>> interest to you.]  QuickFlash's strength continues to be
>>>> high-performance access to raw block data.  However, my desire is to
>>>> have a simple, consistent set of data structures for holding
>>>> interpolated data, and interpolation routines compatible with the
>>>> ghost-cell generation in Paramesh.  (I have not determined how
>>>> Paramesh's inter-block interpolation compares to the method used in
>>>> Fidlr's merge_amr routine that is called by loaddata.)
>>>> 
>>>>    I have been working on some significant functional improvements to
>>>> the QuickFlash library, and I hope to have an interpolation
>>>> infrastructure in the next release.  Unfortunately, development of the
>>>> library is a one-person operation at this point, and is separate from
>>>> the work of my current research position, so I can not make any
>>>> guarantees about how soon it will be available.  (But I can make
>>>> interpolation a priority.)
>>>> 
>>>>    In the meantime, I can certainly provide you with examples of
>>>> filling a uniform mesh with values based on sub-sampling, but they
>>>> would still lead to blocky structure as seen in slices that cover a
>>>> wide range in refinement.  If you wish to develop interpolation
>>>> routines for your own codes using QuickFlash (or help me with
>>>> developing ones within library), I would be more than happy to assist
>>>> you.
>>>> 
>>>> 
>>>> - Nathan
>>>> 
>>>> 
>>>> 
>>>> 
>> 
>>