<div dir="ltr">Hello,<div><br></div><div>The two main requirements for the filter that I am working on are:</div><div>(1) Need to have a GlobalIDs array associated with the nodes</div><div>(2) Single partition per process</div>
<div><br></div><div>If this case satisfies (1) and (2), then I think that the new filter can help.</div><div><br></div><div>Best,</div><div>George</div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Mon, Nov 25, 2013 at 10:28 AM, Utkarsh Ayachit <span dir="ltr"><<a href="mailto:utkarsh.ayachit@kitware.com" target="_blank">utkarsh.ayachit@kitware.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">I'm not surprised piece invariant doesn't solve the problem. Almost<br>
all filters that could use ghost cells to overcome such a problem will<br>
not work since all rely on the reader providing the ghost cells which<br>
you said yours didn't. What is the output type for the dataset from<br>
plot.mir reader? You can use D3 on unstructured datasets to generate<br>
ghost cells -- but that's expensive since it also redistributes the<br>
data. George is currently developing a filter that will simply<br>
generate/communicate ghost cells without redistributing the data.<br>
George, would your filter help in this case, or do I remember things<br>
incorrectly?<br>
<span class="HOEnZb"><font color="#888888"><br>
Utkarsh<br>
</font></span></blockquote></div><br></div>