[Trilinos-Users] AztecOO: result is zero?

[dorian.krause at usi.ch]

On Apr 4, 2011, at 1:06 PM, Sensei wrote:


On Apr 4, 2011, at 12:27pm, dorian.krause at usi.ch<mailto:dorian.krause at usi.ch> wrote:

Hi,

without knowing much details about AztecOO,
it seems to me that this is correct: The example
starts from a random initial guess and converges
to zero (that's a often used test for iterative solvers).


I don't understand... Solving a linear problem with a 2D laplacian matrix should be a piece of cake: shouldn't it converge to the correct result, and not zero?

If the right hand side is zero, the exact solution is zero. In the code,
you can find theses lines:


  Epetra_Vector<http://trilinos.sandia.gov/packages/docs/r6.0/packages/epetra/doc/html/classEpetra__Vector.html> x(Map);
  Epetra_Vector<http://trilinos.sandia.gov/packages/docs/r6.0/packages/epetra/doc/html/classEpetra__Vector.html> b(Map);

  x.Random();


This initializes x with random values and b with zeros (since the zeroOut value is true by default).