[Trilinos-Users] Sparse direct solver and language options for GPU

[Hoemmen, Mark]

On 9/18/14, 12:00 PM, "trilinos-users-request at software.sandia.gov"
<trilinos-users-request at software.sandia.gov> wrote:
>Message: 1
>Date: Thu, 18 Sep 2014 01:17:37 -0700
>From: Tom Anderson <tomacorp at gmail.com>
>To: trilinos-users at software.sandia.gov
>Subject: Re: [Trilinos-Users] Sparse direct solver and language
>	options for	GPU
>Message-ID:
>	<CA+kNXhOs686R2gOc5Jn0Kz3vDPw4Ly_OH2v7DvL2yKWALTF3AQ at mail.gmail.com>
>Content-Type: text/plain; charset="utf-8"
>
>> I am a bit confused by the combination of well-conditioned and
>semi-definite in your description
>I probably didn't use the correct matrix terminology.
>My matrix is a Modified Nodal Analysis circuit solver matrix.
>The only elements I have are resistors, constant current sources, and
>constant voltage sources,
>so it really looks just like example 3 in
>http://www.swarthmore.edu/NatSci/echeeve1/Ref/mna/MNA2.html
>except larger.
>
>I used Anasazi.BasicEigenproblem to find the largest and smallest
>eigenvalues.

Hi Tom -- I should point out that Anasazi doesn't necessarily do what you
think it's doing here, when you ask for the "smallest eigenvalues" (using
the "Which" parameter).  If you really want the "smallest eigenvalues,"
you have to construct a solver for the linear system, and hand it off to
Anasazi as the operator.  Anasazi has some examples to show you how to do
this.  For example,

anasazi/epetra/example/BlockKrylovSchur/BlockKrylovSchurEpetraExGenAmesos.c
pp

has a lot of documentation.  here is a list of some other examples:

http://trilinos.org/docs/dev/packages/anasazi/doc/html/examples.html

mfh