[Paraview] structured to unstructured when using multiblocks filter

Tue Mar 20 14:47:42 EDT 2018

Hi Andy, 

You could find example dataset in the following link, 

[ https://kovan.itu.edu.tr/index.php/s/6RKZuW5rXGS57dR | https://kovan.itu.edu.tr/index.php/s/6RKZuW5rXGS57dR ] 

to work with IndeX volume rendering plugin, i add three grid point ghost region to the tiles in this case but it demonstrated whole extent problem. To write data I am using standard allinputsgrid.py and for volume rendering i am using attached co-processing script (this version creates a band around tiles when I use with standard volume rendering methods of ParaView, i think it is related with three grid ghost point). 

Regards, 

--ufuk 

On 20.03.2018 17:10, Andy Bauer wrote: 

Hi Ufuk, 

If the files are small, could you send them? I'll investigate some. Also, sending your Catalyst Python script may help clue me into what's going wrong. 

Best, 
Andy 

On Mon, Mar 19, 2018 at 4:06 PM, Ufuk Turuncoglu < [ mailto:u.utku.turuncoglu at be.itu.edu.tr | u.utku.turuncoglu at be.itu.edu.tr ] > wrote: 

BQ_BEGIN

Thanks for your valuable comments. Yes, for each model component has two grid 2d and 3d and they are coming from different channels. It is working under 5.4.1 but i'll check the bug and try to fix my version until 5.5 released. BTW, even if I set whole extend (to test it I hardcoded it) in the place that you indicate, the piece and whole extent is same in the files. 

Regards, 

--ufuk 

From: "Andy Bauer" < [ mailto:andy.bauer at kitware.com | andy.bauer at kitware.com ] > 
To: "Ufuk Utku Turuncoglu (BE)" < [ mailto:u.utku.turuncoglu at be.itu.edu.tr | u.utku.turuncoglu at be.itu.edu.tr ] > 
Cc: "paraview" < [ mailto:paraview at paraview.org | paraview at paraview.org ] > 
Sent: Monday, March 19, 2018 3:51:51 PM 
Subject: Re: [Paraview] structured to unstructured when using multiblocks filter 

Hi Ufuk, 

It looks like you're already using multi-channel input as you're adding since you have " g_coprocessorData-> AddInput (strarr[i]); " unless that's always just called once. We're just trying to clarify the "inputs" by naming them channels -- I'm not sure it makes it easier to understand but it makes it easier to talk about with others that understand that concept in Catalyst. 

Setting whole extent would look like the following at line 129 in adaptor.cpp: 

	if (vtkDataSet* grid = vtkDataSet::SafeDownCast (g_coprocessorData-> GetInputDescriptionByName (name)-> GetGrid ())) { 

	ParaViewCoProcessing::ClearFieldDataFromGrid (grid); 
g_coprocessorData-> GetInputDescriptionByName (name)-> SetWholeExtent(ilow, ihigh, jlow, jhigh, klow, khigh); 

	} else { 

Here, *low and *high are global values. For example, for a 1D case if each process has 5 points and you have 3 mpi processes ilow=jlow=jhigh=klow=khigh=0 and ihigh=12. The vtkStructuredGrid::SetExtent() would look like: 
proc 0: 0, 4, 0, 0, 0, 0 
proc 1: 4, 8, 0, 0, 0, 0 
proc 2: 8, 12, 0, 0, 0, 0 

This looks like you're doing it properly in grid.cpp though but I didn't set the SetWholeExtent() in adaptor.cpp and I'm guessing that's what's missing. 

The multi-channel input should work in 5.4.1 but I found a bug that's been fixed in the upcoming 5.5 that was affecting Live. See [ https://gitlab.kitware.com/paraview/paraview/merge_requests/2212 | https://gitlab.kitware.com/paraview/paraview/merge_requests/2212 ] for details on that. It's fairly small so you can probably backport that to your 5.4.1 build if you're more comfortable using 5.4.1. 

Cheers, 
Andy 

On Mon, Mar 19, 2018 at 7:50 AM, Ufuk Utku Turuncoglu (BE) < [ mailto:u.utku.turuncoglu at be.itu.edu.tr | u.utku.turuncoglu at be.itu.edu.tr ] > wrote: 

BQ_BEGIN

Hi Andy, 

Thanks for your help. Yes, each tile has its own grid information. The problem is that if i set whole extent in the adaptor side and write data to disk using allinputsgridwrite.py, the whole extent is same with piece extent. I am not sure what is going on in the background when live visualization activated. If you want to look at it for possible problem, the code is in following link (adaptor.cpp and grid.cpp), 

[ https://github.com/uturuncoglu/RegESM/tree/master/cop | https://github.com/uturuncoglu/RegESM/tree/master/cop ] 

The second option is also worth to try but first i need to look at the example. BTW, i am using 5.4.1. Do you think that the multi-channel input also works with 5.4.1? How it is handled by Catalyst in live mode? Do i have multiple visualization pipeline? 

Regards, 

--ufuk 

On 19.03.2018 13:36, Andy Bauer wrote: 

BQ_BEGIN

Hi Ufuk, 

Currently a multipiece data set always has to be nested under a multiblock dataset. This definitely seems like an unreasonable limitation on a structured grid under a multipiece dataset but I suspect we just haven't had the resources to put into making this work properly. My suggestion would be to try just having your Catalyst adaptor provide a structured grid directly, assuming you only have one grid per mpi process, and then try volume rendering like that. If you do this though, make sure to set the whole extent with vtkCPInputDataDescription::SetWholeExtent(). 

A thought here -- instead of changing your adaptor to generate the structured grid directly instead of a multipiece dataset, maybe create a second "channel" through a second vtkCPInputDataDescription that provides just a structured grid. This way you can provide both dataset types as needed instead of having to modify code back and forth, recompile, etc. to get different functionality. See the Examples/Catalyst/CxxMultiChannelInputExample in the 5.5 source code for some more details on this. 

I can't remember if you were using the editions or just the full ParaView Catalyst build but in 5.5 there will be a volume rendering Catalyst addition. 

Note that the multiblock/multipiece hierarchy will be undergoing a significant change in the near future. 

Best, 
Andy 

On Sun, Mar 18, 2018 at 1:21 PM, Ufuk Turuncoglu < [ mailto:u.utku.turuncoglu at be.itu.edu.tr | u.utku.turuncoglu at be.itu.edu.tr ] > wrote: 

BQ_BEGIN

Hi, 

I am using MPI parallel code under Catalyst co-processing. In this case, i am defining data by creating multi-block dataset and putting multi-piece dataset into the first block (0). In this case, each piece is defined as structured grid. The problem is that, to create volume rendering, i need to use MergeBlocks filter under ParaView but it changes data type from structured grid to unstructured one. I just wonder that is there any way to merge blocks without changing type? Programmable filter etc.? Another workaround could be defining dataset without using multi block under co-processing adaptor code. So, is it possible to represent multi-piece data under co-processing without putting it into multi-bloack dataset. 

Regards, 

--ufuk 

_______________________________________________ 
Powered by [ http://www.kitware.com/ | www.kitware.com ] 

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

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

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

Follow this link to subscribe/unsubscribe: 
[ https://public.kitware.com/mailman/listinfo/paraview | https://public.kitware.com/mailman/listinfo/paraview ] 

BQ_END

BQ_END

BQ_END

BQ_END

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://public.kitware.com/pipermail/paraview/attachments/20180320/dae9a91f/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: state_lr_541_v1_clouds_dvr.py
Type: text/x-python
Size: 12954 bytes
Desc: not available
URL: <https://public.kitware.com/pipermail/paraview/attachments/20180320/dae9a91f/attachment.py>