MantisBT - ParaView
View Issue Details
0006066ParaView(No Category)public2007-11-19 18:282009-05-13 13:46
David Karelitz 
Berk Geveci 
normalfeaturealways
closedfixed 
 
3.6 
0006066: (MARS-high) ParaView is slow with large AMR CTH data
ParaView is very slow when operating on large AMR CTH data. We believe this is related to the number of blocks in the hierarchical dataset (thousands to millions)

This is when performing actions such as threshold, point to cell, contour, and calculator, not when rendering.
No tags attached.
Issue History
2007-11-19 18:28David KarelitzNew Issue
2007-12-06 10:04David KarelitzNote Added: 0009842
2007-12-06 10:04David KarelitzSummary(MARS-normal) ParaView is slow with large AMR CTH data => (MARS-high) ParaView is slow with large AMR CTH data
2008-03-07 07:56Berk GeveciTarget Version => MARS
2008-03-07 08:03Berk GeveciCategory => 3.4
2008-03-07 08:10Berk GeveciStatusbacklog => tabled
2008-03-07 08:10Berk GeveciAssigned To => Berk Geveci
2008-03-11 13:51Ken MorelandNote Added: 0010768
2008-07-24 13:25Utkarsh AyachitCategory3.4 => 3.6
2009-02-16 17:31Ken MorelandNote Added: 0014982
2009-02-16 17:31Ken MorelandStatustabled => @80@
2009-02-16 17:31Ken MorelandResolutionopen => fixed
2009-02-16 20:34Alan ScottNote Added: 0015004
2009-02-16 20:34Alan ScottStatus@80@ => closed
2009-05-13 13:46Utkarsh AyachitTarget VersionMARS => 3.6
2011-06-16 13:10Zack GalbreathCategory => (No Category)

Notes
(0009842)
David Karelitz   
2007-12-06 10:04   
The high-priority filters are:

1. CellToPoint/Dual grid (to enable isosurface and clip by scalar)
2. Calculator (to make the 3 vector components into a vector)
3. Threshold
4. Clip By Scalar
5. Contour
(0010768)
Ken Moreland   
2008-03-11 13:51   
6. Geometry extraction (vtkPVGeometryFilter?).

Recently I used ParaView to load a multiblock data set that took about a minute to get to the point where an image could be created (and only a small fraction of that was taken in rendering) where a simple VTK program that used vtkCompositeDataGeometryFilter was finished in seconds.
(0014982)
Ken Moreland   
2009-02-16 17:31   
I won't go so far as to say this has been completely resolved, but a lot of work has been done to make AMR data sets with large numbers of blocks process faster.

Further improvements will probably have to be done on a filter or component basis. You should raise a separate bug for each thing that needs to be optimized.
(0015004)
Alan Scott   
2009-02-16 20:34   
Bug closed. Pat on the back to the team over the last year!