<div dir="ltr">Humm. What I meant by contiguous is in terms of new global numberings. That is for proc 0: [0, np0), for proc 1: [np0, np1), for proc 2: [np1, np2) ..., for proc p-1: [ np-2, ntotal). I did not mean (topologically) contiguous partitions. Is this not the case?<br>
<br><div class="gmail_quote">On Thu, Apr 12, 2012 at 10:03 AM, Jed Brown <span dir="ltr"><<a href="mailto:jedbrown@mcs.anl.gov">jedbrown@mcs.anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="im"><div class="gmail_quote">On Thu, Apr 12, 2012 at 11:56, Mohammad Mirzadeh <span dir="ltr"><<a href="mailto:mirzadeh@gmail.com" target="_blank">mirzadeh@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
2- Unlike PETSc, ParMetis does not equal number of nodes/elements to each processors. The numbers are close, but not the same as you start partitioning the grid with. It, however, does produce contigious distribution of nodes/elements.</blockquote>
</div><br></div><div>Contiguous partitions are not guaranteed.</div>
</blockquote></div><br></div>