[Paraview] How to extract a subdataset?

[Fred Fred]

Ok, I did not understand "inside out" as "by default inside is out" but rather "clic this if you want that inside be out", I always understood that clipping a space by a given volume was supposed to be an action to limit the working space by the closed volume, as for a view frustum.
Anyway let us comme back to my memory management problem and look at this funny experience.
I have an unstructured dataset with a large number of point data fields.
First I tried to determine the largest dataset I could load in my computer memory. So I increased the number of point data fields till ParaView crashes and I managed to load 1.9GB.
Now another experience. I load the same dataset with 11 scalar fields and 1 vector field, all of this consuming 722MB. Now I clip it with a plane approximatively at the center of the volume. So I would expect that the total memory size necessary for storing the pipeline be in the range 722+722/2 and 722x2, that is between duplicate just half the volume or duplicate it entirely and just keep in mind that there is a cutting plane somewhere. And ParaView crashes for memory reason as just as a litte bit more than half of the job was done.
I do not understand the error message below and especially why ParaView requires 189M elements of 4 bytes each since my dataset has 6.23M points and 6.26M cells, 11 scalar fields and 1 vector field (all as point data, no cell data).
I still think that there is a real issue in memory management.
I use ParaView 3.4.0 on a Mac.

paraview(236,0xa0405720) malloc: *** mmap(size=756289536) failed (error code=12)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
ERROR: In /usr/local/ParaView-3.4.0/VTK/Common/vtkDataArrayTemplate.txx, line 324
vtkIdTypeArray (0x34362930): Unable to allocate 189071391 elements of size 4 bytes. 


From: kmorel at sandia.gov
To: paul.m.edwards at gmail.com; stan1313 at hotmail.fr
CC: paraview at paraview.org
Date: Thu, 3 Sep 2009 09:04:10 -0600
Subject: Re: [Paraview] How to extract a subdataset?





Re: [Paraview] How to extract a subdataset?


Somehow I missed this original email.



I don’t think BuildLinks is a good explanation here.  If you have removed 99 out of 100 cells, I can’t imagine the output growing to take up twice as much memory just because links were built.  The clip filter will also tetrahedral your data, but that should only grow it by a factor of at most 6, not 200.



I assume you are using the Clip filter with Clip Type of Box.  Are you sure that the output contains the cells that are inside the box (the 1%) rather than outside the box (the 99%)?  By default, the clip filter will clip away everything inside the box and leave the rest.  To get only the cells in the box, you have to click on the Inside Out checkbox.  If you set the representation to Surface, do you get what you expect?  Does the cell count drop by a factor of 100?



BTW, using Extract Cells By Region was a good suggestion.  If you are just interested in extracting a group of cells, this filter should be more efficient and consume less memory.



-Ken





On 9/3/09 5:54 AM, "Paul Edwards" <paul.m.edwards at gmail.com> wrote:



I think this is the result of BuildLinks being called.



Regards,

Paul



2009/9/3 Fred Fred <stan1313 at hotmail.fr>

I mean without having the subset be more memory consumming than its including part!

I have an unstructured data set with 9 scalar data fields, which occupies 573 MB, I clip it with a very small box (less than 1% of the whole domain) and the resulting subset occupies 1,920 MB!! Nice!

So:

  1) how is it possible?

  2) how to clip an unstructured dataset with the expected resulting effect on the memory size?





Messenger débarque dans Hotmail ! Essayez-le ! <http://www.windowslive.fr/hotmail/web-messenger/> 



_______________________________________________

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



Follow this link to subscribe/unsubscribe:

http://www.paraview.org/mailman/listinfo/paraview











   ****      Kenneth Moreland

    ***      Sandia National Laboratories

***********  

*** *** ***  email: kmorel at sandia.gov

**  ***  **  phone: (505) 844-8919

    ***      web:   http://www.cs.unm.edu/~kmorel




_________________________________________________________________
Un avatar à votre image ? Créez votre mini-moi !
http://www.ilovemessenger.fr/minimize-me/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20090906/2dcc4b63/attachment-0001.htm>