making new mesh

[grastro]

Hi guys,
I have a set of data which seem to have a wrong mesh in VisIt. my code is supposed to produce meshes in spherical coordinate with r of 256 value, theta of 128 and phi of 64 value for the resolution. but when I open my output data in visit and check the mesh icon, I see a mesh (65*128*256) which is not the same as the code considers. I expect a mesh (256*128*64) instead. what might be wrong here? Is it possible to modify the mesh to have correct numbers of cells? Or I need to define a new mesh?
Thank you in advance

[brugger1]

Hi Grastro,
There isn't a lot you can do in VisIt to change the mesh structure. If it is incorrect, you would most likely need to create a new mesh. Just for a little more context, is the mesh not displaying correctly? What file format are you using?
Eric

[grastro]

Hi Eric,

Sorry for delay, as I was not in my office during these days being involved with some problems. let me explain my problem in more details. I have two data sets. one is in HARM.xmf another one is in coords.h5. Actually, I need to plot velocity vector field from HARM.xmf file and it is not radially directed as it is supposed to be at the first time step (the initial condition of simulation considers a radially directed initial velocity). but I get set of vectors directed approximately towards x axis and is not radially! I also did transformation from spherical to Cartesian (my simulation set up is in spherical coordinates) but I get a set of horizontal vectors exactly directed towards x axis! I'm wondering what is wrong with it!
To find a solution, I mapped data from .xmf file which contains 4-vector velocity, onto a mesh defined in another file which is in coords.h5. this file includes r, theta and phi coordinates with other outputs such as Determinants of metric, etc. the mesh of this output is 65*128*256 while I defined a 256*128*64 mesh in simulation code for r, theta and phi, respectively. so I think the mesh is shown wrong in VisIt, although I could get a set of radially directed velocity vectors when I mapped .xmf data onto .h5 mesh. Of course, I'm not still sure how correct is this procedure!
Anyway, my questions are: 1) why I can not get correct direction (radial) for velocity vectors when I use .xmf mesh while I can plot other scalar quantities like density and pressure correctly (apparently) using its mesh?
2) If mapping onto .h5 file is a correct solution, why its mesh is not the same as it has been defined in simulation code? can I create a new mesh 256*128*64 in .h5 data set to make sure mapping is done correctly?

Regards,
Narges

[brugger1]

Hi Narges,

You mentioned that when you display quantities like density and pressure things look ok (I assume plotted with the Pseudocolor plot). What about when you plot just the mesh with the Mesh plot? Most likely if your pressure and density display correctly, then the mesh is ok. You did note that you mesh is output at 65x128x256 and VisIt thinks it is 256x128x64. I believe the VisIt considers the mesh dimensions in the opposite order from what the xmf file has, so this is mostly expected. I'm not sure about the 64 versus 65 issue though. That could be from a node dimension versus a zone dimension.

Can you be more specific about how you mapped the velocities onto the h5 mesh? Was that through the xmf file or some operation in VisIt? Can you make the data available or a smaller version of your data available? The xmf file would be the more interesting of the two if the size of the data was an issue. For the h5 file, a dump of the header information would probably be useful (h5dump --header mesh.h5 where mesh.h5 is the name of your h5 file), again, if the size of the data was an issue.

Eric

[grastro]

Thanks for the reply. I consider each part separately.

You mentioned that when you display quantities like density and pressure things look ok (I assume plotted with the Pseudocolor plot). What about when you plot just the mesh with the Mesh plot? Most likely if your pressure and density display correctly, then the mesh is ok. You did note that you mesh is output at 65x128x256 and VisIt thinks it is 256x128x64. I believe the VisIt considers the mesh dimensions in the opposite order from what the xmf file has, so this is mostly expected. I'm not sure about the 64 versus 65 issue though. That could be from a node dimension versus a zone dimension.

Yes. The mesh seems correct for .xmf file through which I can plot density and pressure but the velocity vector seems to be in wrong direction (that is not radial, as it should be). I have a mesh with the radius around 1000 which is the same as the radius I've applied for simulation. But when I plot the mesh for coords.h5 (that is supposed to be on the same mesh as .xmf file by simulation set up), it is a cube with x of the value 65, y of 128 and z of 256 values. Through doing transformation from spherical to Cartesian I get a sphere of radius 65 which is not the same as the value I consider for radius in the simulation (around 1000). You can compare these three meshes (one of which is for .xmf file and without transformation while the other two ones are for .h5 file with and without transformation from spherical to Cartesian), in the attachement.

Can you be more specific about how you mapped the velocities onto the h5 mesh? Was that through the xmf file or some operation in VisIt? Can you make the data available or a smaller version of your data available? The xmf file would be the more interesting of the two if the size of the data was an issue. For the h5 file, a dump of the header information would probably be useful (h5dump --header mesh.h5 where mesh.h5 is the name of your h5 file), again, if the size of the data was an issue.

I mapped velocity vector from .xmf file (called as U1U2U3gdet) onto coords.h5 mesh to get a radial direction for it. You can find both data files compressed in two zip files here.

Narges

MeshH5.pdf
MeshH5TransformedFromSphericalToCartisian.pdf
MeshXMF.pdf
HARM_model.xmf.zip
coords.zip
header_info.txt

[brugger1]

I believe your main issue is that you shouldn't be opening the h5 file directly in VisIt. h5 files only know about multidimensional arrays, so when you open it up with VisIt is displays it the best it can which is as a rectangular block. You should only open the xmf file. I suspect that either the velocity components are not correctly referencing things in the h5 file from the xmf file or they are incorrect in the h5 file.

I'll take a look at your files when I have a chance.

[grastro]

Thanks for your comments. I hope I can receive more help as you could take a look at my files. I also try to consider your comments in my efforts.

Narges