[Paraview] paraview display issue on parallel files

Thu Sep 7 16:18:48 EDT 2017

Rilin,

I think I found a solution for you that will work at least most of the time. The solution is to simply write out .vtm files instead of .pvtu files. vtm, which is the multiblock file format, is similar to the pvt* format. The main difference is that when VTK reads in vtm files it keeps the data from different files in different blocks, and that will prevent ParaView from removing what it thinks are internal faces. I’ve attached an example vtkm file to this email.

A bit more explanation and a description of where this will fail. As Andy said the problem is that the files in your dataset have overlapping cells. When ParaView reads in a pvtu file, it combines all the cells from all files into a single mesh, and this mesh will have duplicates of the overlapping cells. By definition, these overlapping cells will share all of their faces. ParaView’s external faces algorithm assumes that if two cells have the same face, that face must be internal and can be removed (which is true in conforming meshes but not when there are duplicate cells). When you first load your mesh into ParaView, everything seems to be working fine because the reader is also creating duplicate points so index-wise the duplicate cells register as different. But many operations (like merge cells, clip, and D3) will find coincident points and merge them together. That’s when ParaView removes all the faces of all duplicate cells.

When ParaView reads the vtm file, it keeps the data from each file in a separate mesh. When ParaView runs algorithms like external faces, it does so independently from each other. So even if the cells are sharing faces, they will be across blocks and will still be considered external. The multiblock structure should remain intact as long as you don’t run a filter that merges them together like D3 or Merge Blocks.

Hope that helps.

-Ken

From: ParaView [mailto:paraview-bounces at paraview.org] On Behalf Of Rilin Shen
Sent: Thursday, September 7, 2017 1:38 PM
To: Bauer, Andy (External Contacts) <andy.bauer at kitware.com>
Cc: paraview at paraview.org
Subject: [EXTERNAL] Re: [Paraview] paraview display issue on parallel files

Dear Andy,

Thanks very much.

You can remap the point variables back onto the points using the Resample with Dataset filter but that won't work for cell data.
I tried the method you advised. Assuming that the original file name is A and the file after Clean to Grid and Clean Cells to Grid is B, I tried two different options for the source for Resample with DataSets and both A and B are not working.  Do I misunderstand something?

I believe that the D3 filter is only available if you have ParaView built/installed with MPI. Even then it will only work if you're running with more than a single MPI process.
I only use serial paraview to process multi-results generated by other programs.  I don't use the parallel paraview to deal with large datasets so I think D3 filter is not suitable for me.

I still suggest writing out the data with either marking the repeated cells as ghost cells or not including them in the file as any other thing you do to get around that can become fairly expensive and manually intensive as you go to bigger datasets.
I would try to not include repeated cells in the original output file. But how to mark the repeated cells as ghost cells? I think there should a vtu/pvtu file format to describe it but I didn't find it on the internet. Could you please share it to me?

BTW,  when I clip a small model with 6700 elements generated by 40 processors, it looks fine. So my question: is there a limit to deal with the large dataset for paraview?

Best regards,

Rilin Shen
Division of Solid Mechanics, School of Astronautics
Harbin Institute of Technology, Harbin 150001, China
Civil Engineering & Engineering Mechanics
Columbia University, New York 10027, U.S.
Tel: (+1) 646-200-2122
Email: shenrilin at gmail.com<mailto:shenrilin at gmail.com>  /  rs3741 at columbia.edu<mailto:shenrilin at 126.com>

2017-09-07 15:00 GMT-04:00 Andy Bauer <andy.bauer at kitware.com<mailto:andy.bauer at kitware.com>>:
Hi,

You can remap the point variables back onto the points using the Resample with Dataset filter but that won't work for cell data.

I believe that the D3 filter is only available if you have ParaView built/installed with MPI. Even then it will only work if you're running with more than a single MPI process.

I still suggest writing out the data with either marking the repeated cells as ghost cells or not including them in the file as any other thing you do to get around that can become fairly expensive and manually intensive as you go to bigger datasets.

Andy

On Wed, Sep 6, 2017 at 3:28 PM, Rilin Shen <shenrilin at gmail.com<mailto:shenrilin at gmail.com>> wrote:
Dear all,

Thanks for your suggestions. It is exactly what you said.

Marking ghost cell would be non-trival work and I tried the second idea.  But only cell and points information is left, other field information associated with points seem to be lost. Is that the case? If that is the case, is there an easy way to attach point data to points again?

BTW, I searched the internet and found that the D3 filter could be a good option for my unstructured grid information. But I cannot find the D3 filter as Filters>Alphabetical>D3. So my second question is how to add D3 filter to my paraview?

Best regards,

Rilin Shen
Division of Solid Mechanics, School of Astronautics
Harbin Institute of Technology, Harbin 150001, China
Civil Engineering & Engineering Mechanics
Columbia University, New York 10027, U.S.
Tel: (+1) 646-200-2122<tel:(646)%20200-2122>
Email: shenrilin at gmail.com<mailto:shenrilin at gmail.com>  /  rs3741 at columbia.edu<mailto:shenrilin at 126.com>

2017-09-06 14:50 GMT-04:00 Andy Bauer <andy.bauer at kitware.com<mailto:andy.bauer at kitware.com>>:
Hi,
Please respond to the entire mailing list so that anyone that wants to participate or follow along can.
I think the issue here is that you have the same cell specified in multiple vtu files and ParaView is getting confused when trying to render them. These are the overlapping cells at the subdomain boundary. My recommendation would be to either mark these repeated cells as ghost cells (see https://blog.kitware.com/ghost-and-blanking-visibility-changes/ for information on that). Other than that, you could use the Clean to Grid filter to get rid of coincident points and then use the Clean Cells to Grid filter to get rid of extra copies of the cell.
Cheers,
Andy

On Tue, Sep 5, 2017 at 6:24 PM, Rilin Shen <shenrilin at gmail.com<mailto:shenrilin at gmail.com>> wrote:

[https://ssl.gstatic.com/docs/doclist/images/icon_10_generic_list.png] Bone.7z<https://drive.google.com/file/d/0B1JLvkeOK5PQVXR3cTBHa2lfQjQ/view?usp=drive_web>
Dear Andy,

Thanks a lot for your help.

As you said, It looks like a ghost cell issue. I attached my dataset. Please check it.

If you need any additional information, please let me know.

Best regards,

Rilin Shen
Division of Solid Mechanics, School of Astronautics
Harbin Institute of Technology, Harbin 150001, China
Civil Engineering & Engineering Mechanics
Columbia University, New York 10027, U.S.
Tel: (+1) 646-200-2122<tel:(646)%20200-2122>
Email: shenrilin at gmail.com<mailto:shenrilin at gmail.com>  /  rs3741 at columbia.edu<mailto:shenrilin at 126.com>

2017-09-05 16:54 GMT-04:00 Andy Bauer <andy.bauer at kitware.com<mailto:andy.bauer at kitware.com>>:
Hi,
This typically happens when ghost cells are marked improperly or the data set has a node/point partitioning and no cells to bridge between the subdomains (though the point partitioning issue usually happens with finite difference simulations instead of FEM simulations). If you could share your dataset it would be easier to diagnose.
Best,
Andy

On Tue, Sep 5, 2017 at 4:23 PM, Rilin Shen <shenrilin at gmail.com<mailto:shenrilin at gmail.com>> wrote:
Dear all,

I'm very happy to meet all of you in this society.

I use a finite element software with multiprocessors(40) to solve problems.

All the .vtu files. associated with each processor are output and then a file .pvtu is generated.

The original file is displayed perfectly as shown in bone.png.  However, when I try to cut it with a plane, it occurs the following issue as clip.png, It seems not be displayed correctly.

Would you please tell me what happened to my data?

Best regards,

Rilin Shen
Division of Solid Mechanics, School of Astronautics
Harbin Institute of Technology, Harbin 150001, China
Civil Engineering & Engineering Mechanics
Columbia University, New York 10027, U.S.
Tel: (+1) 646-200-2122<tel:(646)%20200-2122>
Email: shenrilin at gmail.com<mailto:shenrilin at gmail.com>  /  rs3741 at columbia.edu<mailto:shenrilin at 126.com>

_______________________________________________
Powered by www.kitware.com<http://www.kitware.com>

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView

Search the list archives at: http://markmail.org/search/?q=ParaView

Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20170907/896a06ca/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PUEXsqubar_000001.vtm
Type: application/octet-stream
Size: 2562 bytes
Desc: PUEXsqubar_000001.vtm
URL: <http://public.kitware.com/pipermail/paraview/attachments/20170907/896a06ca/attachment.obj>