I've read some of the other points, and I can easily see one huge bottleneck.<br><br>You've fed STructured Data to Ensight, but processed it (to remove blank cells) into unstructured for Paraview. Then, because of this, you have to run a CellToPoints filter for each isosurface. The way VTK does CellToPoints is with a nice cell averaging routine, which gives you great results but at a huge performance penalty. For a more-balanced test, you should have written out per-point data during the conversion process, to eliminate this reprocessing during the Paraview run.
<br><br>If I've misunderstood you and you actually are using Per-Cell data, then perhaps an Isosurface isn't exactly what you want, but rather a Thresholding (only keep data within a narrow range of your target value). <br>
<br><div><span class="gmail_quote">On 3/21/06, <b class="gmail_sendername">Terry Jordan</b> <<a href="mailto:tejj@hotmail.com">tejj@hotmail.com</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
1) What are you animating? Single Isovalue for multiple timesteps? Multiple<br>isovalues?<br><br>Single IsoValue for multiple timesteps.<br><br><br>2) What data type are you using? Native VTK, Native Ensight, or a 3rd<br>
party. Each of the first two give a somewhat unfair advantage to one<br>product, if you're going for a truly even comparison.<br><br>3rd party cfd data. We have written readers for paraview and ensight for<br>this data.
<br><br>3) What data format are you using? Structured, unstructured? 2d, 3d?<br>For ensight it is structured, and in paraview we are using unstructured.<br><br>4) What pipeline/viz tools are you using?<br>I'm not sure what you are asking here. But I will guess you mean vtk
<br>pipeline in paraview.<br>In paraview it's Reader->CellToPoint->Contour.<br>and in ensight it is just Reader->IsoSurface.<br><br>5) What exactly are you counting time for? Data load, or just computation?<br>I bet you're just timing the actual time to generate all the isosurfaces,
<br>and Ensight gets a big win because it preloads all the data at once where<br>Paraview loads it as it goes.<br>We are counting the time it takes to generate all of the isosurfaces.<br><br>Another problem could be that the data is taken as a structured grid in
<br>ensight and an unstructured grid in paraview.<br><br>----Original Message Follows----<br>From: "Randall Hand" <<a href="mailto:randall.hand@gmail.com">randall.hand@gmail.com</a>><br>To: "Terry Jordan" <
<a href="mailto:tejj@hotmail.com">tejj@hotmail.com</a>><br>CC: <a href="mailto:paraview@paraview.org">paraview@paraview.org</a><br>Subject: Re: [Paraview] Ensight<br>Date: Tue, 21 Mar 2006 10:21:08 -0600<br><br>Lots of information missing here.
<br>1) What are you animating? Single Isovalue for multiple timesteps? Multiple<br>isovalues?<br>2) What data type are you using? Native VTK, Native Ensight, or a 3rd<br>party. Each of the first two give a somewhat unfair advantage to one
<br>product, if you're going for a truly even comparison.<br>3) What data format are you using? Structured, unstructured? 2d, 3d?<br>4) What pipeline/viz tools are you using?<br>5) What exactly are you counting time for? Data load, or just computation?
<br>I bet you're just timing the actual time to generate all the isosurfaces,<br>and Ensight gets a big win because it preloads all the data at once where<br>Paraview loads it as it goes.<br><br><br>On 3/21/06, Terry Jordan <
<a href="mailto:tejj@hotmail.com">tejj@hotmail.com</a>> wrote:<br> ><br> > We are running some test of the performance of ensight versus<br> > paraview. In<br> > the test we use the same data set and create an iso surface
<br> > (contour). Then<br> > we animate the transient data set. We are seeing about 1 order of<br> > magnitude<br> > increase in performance over paraview (about 2.5 mins to 29 mins). Does<br> > anyone know if ensight is doing something to speed up the performance
<br>such<br> > as decimation or are their algorithms that much faster?<br> ><br> > Thanks in advance.<br> ><br> ><br> > _______________________________________________<br> > ParaView mailing list<br> >
<a href="mailto:ParaView@paraview.org">ParaView@paraview.org</a><br> > <a href="http://www.paraview.org/mailman/listinfo/paraview">http://www.paraview.org/mailman/listinfo/paraview</a><br> ><br><br><br><br>--<br>Randall Hand
<br>Visualization Scientist,<br>ERDC-MSRC Vicksburg, MS<br>Homepage: <a href="http://www.yeraze.com">http://www.yeraze.com</a><br><br><br>_______________________________________________<br>ParaView mailing list<br><a href="mailto:ParaView@paraview.org">
ParaView@paraview.org</a><br><a href="http://www.paraview.org/mailman/listinfo/paraview">http://www.paraview.org/mailman/listinfo/paraview</a><br><br><br></blockquote></div><br><br clear="all"><br>-- <br>Randall Hand<br>Visualization Scientist,
<br>ERDC-MSRC Vicksburg, MS<br>Homepage: <a href="http://www.yeraze.com">http://www.yeraze.com</a>