[Trilinos-Users] Epetra_CrsMatrix::InsertGlobalValues

[Heroux, Michael A]

Antonella,

The interpretation of having multiple instances of a nonzero value is that the entries should have an additive effect.  Since these are global objects, I think this is the only reasonable way to interpret multiple instances of a value.

If you want the 1.0 value simultaneously on two or more processors, this suggests that you are doing some kind of ghosting operation, which is typically support through other means than replicating values.

Mike

From: Antonella Longo <antonella.longo at ingv.it<mailto:antonella.longo at ingv.it>>
Date: Sunday, March 8, 2015 at 3:05 AM
To: "trilinos-users at software.sandia.gov<mailto:trilinos-users at software.sandia.gov>" <trilinos-users at software.sandia.gov<mailto:trilinos-users at software.sandia.gov>>
Subject: [EXTERNAL] [Trilinos-Users] Epetra_CrsMatrix::InsertGlobalValues

I insert a fixed number (1.0) in a specific position (row1,col1) of
an Epetra_CrsMatrix with the function InsertGlobalValues.
The problem is that more than one PID owns the (row1,col1) position,
so that when i call the function InsertGlobalValues for the (row1,col1) position
each PID gives its contribution. As a result, after the fillComplete call
all these contributions will sum up, resulting in a number greater than 1
in position (row1,col1).
I also tried with the ReplaceGlobalValues function, but also these contributions
are summed up with the fillComplete call.

Is there any way to prevent this problem?

Antonella


--
-----------------------------------------------
Antonella Longo
Istituto Nazionale di Geofisica e Vulcanologia
Sezione di Pisa
Via Uguccione della Faggiola, 32
Pisa, Italy
email: antonella.longo at ingv.it<mailto:antonella.longo at ingv.it>
Tel. office: 050 8311939
Cell.: 339 7532089
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://software.sandia.gov/pipermail/trilinos-users/attachments/20150308/b020696d/attachment.html>