Timo and Wolfgang,
Thanks for the response. Here's what I find by working with step-40.
> > viz. PETSC ERROR: Mat mat,Vec y: local dim 4160 8450 (on two processes).
>
> What are the (global) vector and matrix sizes supposed to be? Which
> parts of the matrix and vectors are supposed to be stored locally? What
> I read here is that one of the vectors is partitioned differently than
> the matrix...
I get the same behaviour from my application as step-40; and the same
behaviour with completely_distributed_solution and
locally_relevant_solution (error message).
The discussion is now only about errs from step-40.
The .size () for vectors and .m () and .n () for matrices all report
matrices and vectors have the same global size; 4225. This I expected.
The error is reproduced by adding the line
const double norm = system_matrix.matrix_norm_square
(completely_distributed_solution);
right afer solver.solve () in step-40.
Then, PETsc complains that the objects are Nonconforming object sizes:
Mat mat, Vec y: local dim 2080 4225! (two processes)
Mat mat, Vec y: local dim 1089 4225! (four processes)
Which I read in a similar way as you did Wolfgang. For me, PETSc knows
these are distributed objects, but somehow the local dim of the vectors is
being given incorrectly (getting global instead), or, these vectors are
actually not being distributed / partitioned and still have global size. I
doubt the last one...
The n_locally_owned_dofs queried for vectors and matrices match equally
on each processor. How MatMult is getting it wrong?
Timo: Was trying the MatMult thing with completely_distributed_solution
what you meant? I get the same problem :-|
Best,
Toby
-----
Toby D. Young
Assistant Professor
Institute of Fundamental Technological Problems
Polish Academy of Sciences
ul Adolfa Pawinskiego 5b
02-106 Warsaw
Poland
www: http://www.ippt.gov.pl/~tyoung
skype: stenografia
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
