[overture] Re: v24 PETSc convergence
- From: "Ale" <aorchini@xxxxxxxxxx>
- To: <overture@xxxxxxxxxxxxx>
- Date: Wed, 6 Jun 2012 17:38:15 +0200
Yes,
I forgot to report, in my case is not a fatal error, also v24 is going on
integrating. but I’m going to fix it.
Thanks Bill,
Regards,
Alessandro
From: Bill Henshaw
Sent: Wednesday, June 06, 2012 5:21 PM
To: overture@xxxxxxxxxxxxx
Subject: [overture] Re: v24 PETSc convergence
Hi Alessandro,
It turns out that I was not checking the error return from PETSc in v23
so the difference is that in v24 the error return is now being checked
and the message below is printed. Sometimes the error is not fatal.
Otherwise you may need to increase the max number of iterations
or increase the number if ILU levels, for example.
...Bill
On 06/06/2012 03:03 AM, Alessandro Orchini wrote:
Hi,
I'm trying to re-run with v24 some old simulations that I already did with
v23, (I'm using same commands both for ogen and cgins, but running them with
v24 version of the programs)
and I'm facing a PETSc error which I'm sure, is not showed up in v23 with the
same parameters:
PETScEquationSolver:ERROR: Solution diverged! reason=-3:
KSP_DIVERGED_NULL = -2,
KSP_DIVERGED_ITS = -3,
KSP_DIVERGED_DTOL = -4,
KSP_DIVERGED_BREAKDOWN = -5,
KSP_DIVERGED_BREAKDOWN_BICG = -6,
KSP_DIVERGED_NONSYMMETRIC = -7,
KSP_DIVERGED_INDEFINITE_PC = -8,
SP_DIVERGED_NAN = -9,
KSP_DIVERGED_INDEFINITE_MAT = -10
NOTE 1: to see more information turn on the '-info' PETSc option (e.g. in
your .petscrc)
NOTE 2: to avoid the divergence error '-4' you can set the Oges option
'maximum allowable increase in the residual'
Is this a new divergence-control inserted in the new version? And why the
same simulation in v23 does not diverge?
Thanks!
Other related posts:
