IDACalcIC unable to recover

From: Bernhard Barkow <bb@...>
To: sundials-users@...
Content-Type: text/plain; charset=WINDOWS-1252; format=flowed; delsp=yes
Mime-Version: 1.0 (Apple Message framework v936)
Subject: IDACalcIC unable to recover
Date: Thu, 17 Sep 2009 17:47:56 +0200
X-Mailer: Apple Mail (2.936)
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to
8bit by nsplists.llnl.gov id n8HFm0o5017086

Hi,

while trying to integrate a simple PDE system with IDAS (from SUNDIALS
2.4.0), using the dense linear solver, I encountered the following
problem:
--
[IDAS ERROR]  IDACalcIC
    The residual routine or the linear setup or solve routine had a
recoverable error, but IDACalcIC was unable to recover.

IDACalcIC: -14
--

According to the documentation, this happens when "…
- the residual function,
- or the linear solver’s setup
- or solve function
had a recoverable error, but IDACalcIC was unable to recover."

My residual function returns 0; as far as I found out, the dense
linear solver also returns 0, so I guess it must be about the solver's
setup function; searching the list archives, I found that this might
be due to my Jacobian being singular, or a problem in the LU
factorization.
I do not provide an explicit Jacobian but have IDAS compute it
numerically; computing it manually using perturbations in y and yp
yields a non-singular matrix.
And for completenes' sake, IDADlsGetLastFlag() returns 0 (called after
IDACalcIC()).

As for the system itself, it is a discretized rope, consisting of
several nodes (> 2), the end points of which are fixed. Accordingly, I
define the equations concerning the boundary nodes as algebraic (using
IDASetId() - is that approach correct?).

Hopefully somebody can provide some insight…

Regards,
Bernhard Barkow

From: Bernhard Barkow <bb@...>
To: sundials-users@...
In-Reply-To: <7vmqap$1lfrkn@...>
Content-Type: text/plain; charset=WINDOWS-1252; format=flowed; delsp=yes
Mime-Version: 1.0 (Apple Message framework v936)
Subject: Re: [sundials-users] IDACalcIC unable to recover
Date: Fri, 18 Sep 2009 16:42:10 +0200
References: <7vmqap$1lfrkn@...>
X-Mailer: Apple Mail (2.936)
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to
8bit by nsplists.llnl.gov id n8IEgKKL014639

On 2009-09-18, at 15:06, by way of Carol Woodward wrote:
 > From: Bernhard Barkow <bb@...>
 > To: sundials-users@...
 >
 > Hi,
 >
 > while trying to integrate a simple PDE system with IDAS (from SUNDIALS
 > 2.4.0), using the dense linear solver, I encountered the following
 > problem:
 > --
 > [IDAS ERROR]  IDACalcIC
 >   The residual routine or the linear setup or solve routine had a
 > recoverable error, but IDACalcIC was unable to recover.
 >
 > IDACalcIC: -14
 > --
 >
 > According to the documentation, this happens when "…
 > - the residual function,
 > - or the linear solver’s setup
 > - or solve function
 > had a recoverable error, but IDACalcIC was unable to recover."
 >
 > My residual function returns 0; as far as I found out, the dense
 > linear solver also returns 0, so I guess it must be about the solver's
 > setup function; searching the list archives, I found that this might
 > be due to my Jacobian being singular, or a problem in the LU
 > factorization.
 > I do not provide an explicit Jacobian but have IDAS compute it
 > numerically; computing it manually using perturbations in y and yp
 > yields a non-singular matrix.
 > And for completenes' sake, IDADlsGetLastFlag() returns 0 (called after
 > IDACalcIC()).
 >
 > As for the system itself, it is a discretized rope, consisting of
 > several nodes (> 2), the end points of which are fixed. Accordingly, I
 > define the equations concerning the boundary nodes as algebraic (using
 > IDASetId() - is that approach correct?).
 >
 > Hopefully somebody can provide some insight…


Hi again,

sorry to bother anybody, I "solved" the problem (as I had suspected,
there was a rather dumb error in my residual function; I just had
looked at it so often that I was unable to find it, and was at my
wits' end).

Regards,
Bernhard


____________________________________________________
_________________________________ Bernhard Barkow __
___________________ gpg key ID _ A89F09C45921020D __