[Trilinos-Users] CrsMatrix behavior

[Hysom, David A.]

Hi,


I'm using Tpetra::Import to construct a matrix with a

2D processor layout from an existing 1D layout.

This works, however, there's a problem in that

some processors end up with empty rows (their

portion of a global row). This is causing LOBPCG

to fail to converge.


If I build a second 2D matrix but leave out the empty

rows (I mean, in the map), LOBPCG works as expected.


Is this a known issue; is it a feature; is it a bug?


thanks, David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://trilinos.org/pipermail/trilinos-users/attachments/20170302/8dbaa601/attachment.html>