dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

Anders Logg wrote:
> On Fri, Apr 23, 2010 at 09:34:16AM +0100, Garth N. Wells wrote:
>> This sounds very PETSc-speecific. Should it be a parameter of
>> PETScMatrix only?
>>
>> Garth
> 
> Yes, perhaps. I find it a bit strange that MatZeroRows breaks for
> rows without elements.
>

Send a PETSc bug report?

A solution would be to check the number of row entries, if possible.

Garth


> The problem with making it a PETScMatrix-specific option is it would
> be very inconvenient to set the option. One would need to cast the
> Matrix to a PETScMatrix and then set the option, or just work with
> PETScMatrix in which case the code breaks when changing backend.
> 
> How about making it a global PETSc option?
> 
>   dolfin::parameters["petsc"]["use_mat_zero_rows"]
> 
> --
> Anders
> 
> 
>> -------- Original Message --------
>> Subject: [Branch ~dolfin-core/dolfin/main] Rev 4695: Add parameter
>> "use_ident"	in DirichletBC. Default value is true (same
>> Date: Thu, 22 Apr 2010 15:17:22 -0000
>> From: noreply@xxxxxxxxxxxxx
>> Reply-To: noreply@xxxxxxxxxxxxx
>> To: Garth Wells <gnw20@xxxxxxxxx>
>>
>> ------------------------------------------------------------
>> revno: 4695
>> committer: Anders Logg <logg@xxxxxxxxx>
>> branch nick: dolfin-dev
>> timestamp: Thu 2010-04-22 17:14:50 +0200
>> message:
>>   Add parameter "use_ident" in DirichletBC. Default value is true (same
>>   behavior as earlie). If set to false, the matrix will be manipulated
>>   using setitem instead (useful when PETSc complains about missing entries
>>   in row).
>> modified:
>>   ChangeLog
>>   dolfin/fem/BoundaryCondition.h
>>   dolfin/fem/DirichletBC.cpp
>>   dolfin/fem/DirichletBC.h
>>   dolfin/la/GenericMatrix.cpp
>>   dolfin/la/KrylovSolver.cpp
>>   dolfin/la/KrylovSolver.h
>>   dolfin/nls/NewtonSolver.cpp
>>
>>
>