[Paraview] Paraview performance

[Alberto Fernandez]

I'm using the clip filter.
The reason for using ensigth is the support for time steps. It seems a 
lot harder to create the same thing in vtk format. The calculation also 
includes mesh deformation and it seemed easier in ensigth.
I'll check the log to see what is taking so long. However, with the big 
model I can not even open it, it´s taking 8 hours now.

Berk Geveci escribió:
> One way of pinpointing  the bottleneck is to look at Window -> Timer 
> Log. I am guessing IO is the problem here. However, clip filter might 
> be also taking a long time with 4.5 million cells on 1 processor.  I 
> recommend switching to at least binary. A 4.5 Gb ascii file is 
> ridiculous. A format that supports compression like the VTK-XML 
> formats might also help. Also, are you using the cut or clip filter?
>
> -berk
>
> On 10/16/06, * Alberto Fernandez* <alberto at cognitdesign.com 
> <mailto:alberto at cognitdesign.com>> wrote:
>
>     Even for a simple solid rendering. I'm trying to open a case with 4.5
>     million and its been working for 6 hours now. Is it normal to have
>     that
>     kind of performance if the data is in ascii? Granted the geo file
>     is big
>     (4Gb) but I thought it could be handled.
>     I'm going to check out HAVS. Thanks for the suggestion.
>
>     Claudio Silva escribió:
>     > What kind of rendering are you doing?
>     >
>     > For volume rendering (and isosurface generation), HAVS can
>     render up
>     > to 6 million tets/sec on the latest graphics hardware (single
>     machine).
>     >
>     > HAVS has recently been added to Paraview, and is also available
>     stand
>     > alone:
>     >
>     > havs.sourceforge.net <http://havs.sourceforge.net>
>     >
>     >
>     > On Oct 15, 2006, at 11:06 AM, Alberto Fernandez wrote:
>     >
>     >> I have a question regarding paraview performance. I'm using
>     paraview
>     >> to visualize CFD results. The results are written in ensight
>     format
>     >> as ascii. I have been doing some work on a model with a
>     unstructured
>     >> mesh of about 1.5 million cells. Using simply a clip filter cutting
>     >> with a scalar takes a lot of time. It is not big for just a
>     frame but
>     >> we are calculating 20 time steps. If we want to calculate that
>     >> cliping (to find a free surface) to create an animation it takes
>     >> about 15 or 20 minutes. The machine is an AMD athlon with  4 GB.
>     >> We are interested in processing a calculation with around 5 million
>     >> cells. We have tried and it takes hours.
>     >> Is there a way of improving this performance through some
>     settings or
>     >> similar?
>     >> Does having the results written in ascii (it takes quite some room)
>     >> have some effect on performance?
>     >>
>     >> --Alberto Fernández
>     >> alberto at cognitdesign.com <mailto:alberto at cognitdesign.com>
>     >>
>     >> Cognit Design
>     >> C/Arquitecte Sert, 31 Local 3    Vila Olímpica
>     >> 08005  Barcelona
>     >> Telf: 93 532 29 94
>     >> Fax: 93 531 59 42
>     >>
>     >> _______________________________________________
>     >> ParaView mailing list
>     >> ParaView at paraview.org <mailto:ParaView at paraview.org>
>     >> http://www.paraview.org/mailman/listinfo/paraview
>     >
>
>     --
>     Alberto Fernández
>     alberto at cognitdesign.com <mailto:alberto at cognitdesign.com>
>
>     Cognit Design
>     C/Arquitecte Sert, 31 Local 3
>     Vila Olímpica
>     08005  Barcelona
>     Telf: 93 532 29 94
>     Fax: 93 531 59 42
>
>     _______________________________________________
>     ParaView mailing list
>     ParaView at paraview.org <mailto:ParaView at paraview.org>
>     http://www.paraview.org/mailman/listinfo/paraview
>
>

-- 
Alberto Fernández
alberto at cognitdesign.com

Cognit Design
C/Arquitecte Sert, 31 Local 3    
Vila Olímpica
08005  Barcelona
Telf: 93 532 29 94
Fax: 93 531 59 42