<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Natalie Happenhofer wrote:
<blockquote cite="mid:BAY122-W2528F2C72D690677F4DAB3BCA40@phx.gbl"
type="cite">
<style>
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
font-size: 10pt;
font-family:Verdana
}
</style>Ok, thanks a lot, I believe I understand.. <br>
<br>
But that brings me to another question: <br>
I´m running paraview on 8 nodes, and my data is sliced into 8 pieces,
which I can see when it´s rendering or with the ProcessId-Filter. In my
case, its a cuboid and it´s divided into 8 pieces, sliced once in every
layer. In order to make the parallelism useful for the algorithmus I
implemented in my filters, I need to change the domain for every
processor, i.e I want to slice the data into 8 pieces, but slicing it 7
times in the xy-layer. How can I achieve this, where are the domains
for each processor specified?<br>
</blockquote>
Paraview will normally use an Extent Translator which will sub-divide
in each direction alternatively, as you have noticed.<br>
<br>
You can override thsi default behavior with two lines of code in your
reader/filter. See further documentation for class vtkExtentTranslator<br>
<i><br>
vtkStreamingDemandDrivenPipeline *exec =
static_cast<vtkStreamingDemandDrivenPipeline*>(this->GetExecutive());<br>
exec->GetExtentTranslator(outInfo)->SetSplitModeToXSlab();</i><br>
<br>
<br>
Jean --<br>
Swiss National Supercomputing Center<br>
</body>
</html>