[visit-developers] avtRectilinearDomainBoundaries in Silo Plugin?

Cyrus D. Harrison cyrush at llnl.gov
Mon Oct 27 14:11:58 EDT 2008


To Summarize: We don't have a flexible way to connect connectivity info 
with a particular mesh. Conn-info in our cache appears
to always be linked to "any_mesh" & VisIt respects this convention 
implicitly.

For the Silo Plugin: When I am constructing the domain boundaries object 
I need to know the mesh type - however there is no real link between
the mesh & the conn-info (I don't even have a mesh name to go by). My 
current approach is to look at the root TOC find the first
mulitmesh with > 0 nblocks and use its type. This appears to work fine - 
I am going to run it through the test suite and see if
there are any hiccups.

-Cyrus

Eric Brugger wrote:
>
> Mark,
>
> When I added the use of avtRectilinearDomainBoundaries to the XDMF 
> reader I ran into issue you are talking about. I had to name it 
> "any_mesh", when I gave it the actual name of the mesh it didn't work. 
> At the time I looked at the other readers to see what they did and all 
> but one used "any_mesh". My guess is that the one that doesn't use 
> "any_mesh" doesn't work. So I think this is something that should be 
> straightened out at some point.
>
> Eric
>
> At 07:09 AM 10/20/2008, Mark Miller wrote:
>> Hi Cyrus,
>>
>> I would think the selection could be made on the basis of the meshtype
>> of the first, non-empty block of a given multimesh.
>>
>> Also, I note that AUXILIARY_DATA_DOMAIN_BOUNDARY_INFORMATION is cached
>> with "any_mesh" in Silo plugin and that may need to be changed to
>> include the proper mesh name. In fact, in its present form, how would it
>> work if indeed we hit a silo file with a mesh whose name was 'any_mesh'?
>>
>> Mark
>>
>> On Sun, 2008-10-19 at 17:58, Cyrus Harrison wrote:
>> > Hank - Thanks for verifying this for me. All the connectivity info is
>> > there, we just need a sensible way to select between rectilinear &
>> > curvilinear. Here is a another observation/question hybrid - While
>> > Silo can store several multi meshes it looks like our plugin supports
>> > a single decomposition scheme? Probably as a side effect of this we
>> > don't have  anything that links the connectivity info to a particular
>> > mesh. This is what I went searching for to resolve the rect vs curvi
>> > domain boundaries, in a general way but I came up empty handed. If
>> > anyone knows something obvious I overlooked please fill me in.
>> >
>> > So to resolve the domain boundary issue would reading the first
>> > multimesh and checking its mesh type be acceptable, or is this to much
>> > of a hack?
>> >
>> > Also - Eric I would like to get this fix into the next release, but
>> > since it is a very small change I can do it on the RC.
>> >
>> > -Cyrus
>> >
>> >
>> >
>> > On Sat, Oct 18, 2008 at 7:06 AM, Hank Childs <hankchilds at yahoo.com> 
>> wrote:
>> > > Hi Cyrus,
>> > >
>> > > I looked over the code and have concluded that, as you suspected, 
>> the Silo reader does not properly handle this case.  It only 
>> instantiates an avtCurvilinearDomainBoundaries object.  For the case 
>> you are describing, as you figured out, it should be making an 
>> avtRectilinearDomainBoundaries object.  So the Silo reader needs to 
>> be enhanced to see what the mesh type is and create the right kind of 
>> object from it.
>> > >
>> > > -Hank
>> > >
>> > >
>> > > --- On Fri, 10/17/08, Cyrus D. Harrison <cyrush at llnl.gov> wrote:
>> > >
>> > >> From: Cyrus D. Harrison <cyrush at llnl.gov>
>> > >> Subject: [visit-developers] avtRectilinearDomainBoundaries in 
>> Silo Plugin?
>> > >> To: "VisIt Developers" <visit-developers at email.ornl.gov>
>> > >> Date: Friday, October 17, 2008, 4:58 PM
>> > >> Hi Everyone,
>> > >> I was able to unwind the nodelists in Tom's Rectilinear
>> > >> Silo MMAdj
>> > >> object example to get things rolling but I hit another road
>> > >> block.
>> > >>
>> > >> I verified that the facelist filter is correctly removing
>> > >> the internal
>> > >> faces between the domains -  but when I try to use ghost
>> > >> zones it crashes.
>> > >> It is crashing while exchanging domain boundary info - the
>> > >> avtCurvilinearDomainBoundaries object is being used but my
>> > >> data is
>> > >> rectilinear.
>> > >>
>> > >> In avtCurvilinearDomainBoundaries::ExchangeMesh it tries to
>> > >> cast the
>> > >> datasets to vtkStructuredGrids, but they are of course
>> > >> vtkRectilinearGrids.
>> > >> (I verified this using dynamic_casts). I looked in
>> > >> avtSiloFileFormat and
>> > >> I didn't see any path currently creating
>> > >> avtRectilinearDomainBoundaries
>> > >> - am I overlooking something there? Are there any
>> > >> multi-domain
>> > >> rectilinear silo example data files w/ ghosts?
>> > >> Thanks,
>> > >> -Cyrus
>> > >
>> > > __________________________________________________
>> > > Do You Yahoo!?
>> > > Tired of spam?  Yahoo! Mail has the best spam protection around
>> > > http://   mail.yahoo.com
>> > >
>> -- 
>> Mark C. Miller, Lawrence Livermore National Laboratory
>> email: mailto:miller86 at llnl.gov
>> (M/T/W) (925)-423-5901 (!!LLNL BUSINESS ONLY!!)
>> (Th/F)  (530)-753-8511 (!!LLNL BUSINESS ONLY!!)
>
>
>



More information about the visit-developers mailing list