<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 07/22/2012 05:14 PM, Jed Brown wrote:
<blockquote
cite="mid:CAM9tzS=8zYgEuHBob4hzd-maU3kq77U4vc3aDbCN0ZjwJW5bDw@mail.gmail.com"
type="cite">
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<div class="gmail_quote">On Sun, Jul 22, 2012 at 9:49 AM, Umut
Tabak <span dir="ltr"><<a moz-do-not-send="true"
href="mailto:u.tabak@tudelft.nl" target="_blank">u.tabak@tudelft.nl</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
Dear all,<br>
<br>
I am testing some iterative methods with MATLAB and aside with
PETSc however I have a question which might be answered in the
documentation however I could not find that?<br>
<br>
In MATLAB, at least on recent versions, one can specify a drop
tolerance for the incomplete cholesky preconditioner, I was
wondering if the same is possible with PETSc or not?<br>
</blockquote>
<div><br>
</div>
<div>You can get an ILUT using MatSuperluSetILUDropTol
(-mat_superlu_ilu_droptol) or from Hypre's PILUT
(-pc_hypre_pilut_tol)..</div>
</div>
</blockquote>
Dear Jed Thanks for this tip, I will try...<br>
<blockquote
cite="mid:CAM9tzS=8zYgEuHBob4hzd-maU3kq77U4vc3aDbCN0ZjwJW5bDw@mail.gmail.com"
type="cite">
<div class="gmail_quote">
<div><br>
</div>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
One more question, is the condition number order
1e+6(estimated with condest in MATLAB) rather high for an
iterative method? With icc, with a drop tolerance of 1e-3 or
1e-4, as a preconditioner to pcg, I can get decent iteration
numbers to convergence in MATLAB, it is sometimes even faster
than solving the system with the available factorization
information and I was wondering if I can make it faster with
some other options in PETSc or not?<br>
</blockquote>
<div><br>
</div>
<div>What continuum equations are you solving? What
discretization?</div>
</div>
</blockquote>
Helmholtz equation, 3d discretization of a fluid domain, basically
the operator is singular however for my problem I can delete one of
the rows of the matrix, for this case, I and get a non-singular
operator that I can continue my operations, basically, I am getting
a matrix with size n-1, where original problem size is n. However,
this application is pretty problem specific, then I can use this
full-rank matrix in linear solutions. The condition number estimate
belongs to this full-rank matrix that is extracted from the original
singular operator...<br>
<br>
</body>
</html>