MantisBT - ParaView
View Issue Details
IDProjectCategoryView StatusDate SubmittedLast Update
0012648ParaView(No Category)public2011-10-12 18:312012-10-29 17:04
ReporterAlan Scott 
Assigned ToGeorge Zagaris 
PriorityurgentSeverityminorReproducibilityhave not tried
StatusclosedResolutionfixed 
PlatformOSOS Version
Product Version3.12 
Target VersionFixed in Version3.98.0 
ProjectSandia
Topic Name
Typeincorrect functionality
Summary0012648: CTH AMR does not work remote server
DescriptionParaView 3.12.0 RC-2. CTH AMR appears broken remote server.

Linux client. Remote server.

Open Dave's small CTH. OK. Change to surface. Nothing shows up. This is the bug.

Filters/ Alphabetical/ Clean To Grid. Apply. Now you can see the object.

You can see the CTH AMR dataset local server, so I believe this one is a remote server issue.

Steps To Reproduce
Additional Information
TagsNo tags attached.
Relationships
Attached Files
Issue History
Date ModifiedUsernameFieldChange
2011-10-12 18:31Alan ScottNew Issue
2011-10-13 10:44Utkarsh AyachitAssigned To => George Zagaris
2011-10-13 15:54David PartykaNote Added: 0027573
2011-10-14 12:04George ZagarisNote Added: 0027576
2012-04-23 16:53George ZagarisNote Added: 0028519
2012-04-23 21:29George ZagarisNote Added: 0028522
2012-04-23 21:29George ZagarisStatusbacklog => gatekeeper review
2012-04-23 21:29George ZagarisResolutionopen => fixed
2012-04-24 14:09Utkarsh AyachitStatusgatekeeper review => customer review
2012-04-30 21:55Alan ScottNote Added: 0028560
2012-04-30 21:55Alan ScottStatuscustomer review => closed
2012-04-30 21:55Alan ScottFixed in Version => git-master
2012-10-29 17:04Utkarsh AyachitFixed in Versiongit-master => 3.98.0

Notes
(0027573)
David Partyka   
2011-10-13 15:54   
Confirmed and recreated.
(0027576)
George Zagaris   
2011-10-14 12:04   
We are able to reproduce this locally. The issue is essentially that CTH AMR is a little different than the Berger-Collela AMR that the latest changes were assuming. CTH AMR is using the ghost layers to figure out internal block faces and exclude them from rendering. The Berger-Collela AMR is using the AMR box at the root level which covers the entire domain to figure out which blocks & block faces to process for rendering the surface. Consequently, since with CTH AMR there may not be any data at level 0 (due to refinement) the AMRbox at level 0 was not computed correctly and all other blocks would fall outside of it and hence not be processed.

This has been fixed in the "amr-enhancements" branch on stage. I added an ivar to the vtkHierarchicalBoxDataSet that defines whether the data is CTH AMR and is being set within the vtkSpyPlotReader. Then, in the PV Geometry filter, if the data is CTH AMR, it will process them as regular CompositeDataSet. Otherwise, it will process them as Berger-Collela AMR.

I would like some feedback on this. Is this the right way to handle this. Albeit this approach works, it seems that we are using the same data-structure in two different roles, does it make sense to create a more specialized data-structure for Berger-Collela AMR, i.e., extend from vtkHierarchicalBoxDataSet ?
(0028519)
George Zagaris   
2012-04-23 16:53   
This should no longer be an issue. The AMR data-structures have been split to vtkOverlappingAMR which implements Berger-Collela AMR and vtkNonOverlappingAMR which is compatible with CTH.
(0028522)
George Zagaris   
2012-04-23 21:29   
Addressed when merged AMR-Refactoring branch
(0028560)
Alan Scott   
2012-04-30 21:55   
Tested remote server, Linux, master.