[petsc-dev] MatCreateTranspose semantics

[Jeff Hammond]

I'm still unable to get a basic matrix transpose working.  I may be stupid,
but I cannot figure out why the object is in the wrong state, no matter
what I do.

This is the full code:

https://github.com/jeffhammond/PRK/commit/617973dfbe07d64cc2c0418a1702c418d51802c5

jrhammon at jrhammon-nuc:~/PRK/C1z$ make petsc && ./transpose-petsc -i 10 -n
100
make: Nothing to be done for 'petsc'.
Parallel Research Kernels version 2020
C11/PETSc Transpose: B += A^T
Number of processes  = 1
Number of iterations = 10
Matrix order         = 100
[0]PETSC ERROR: --------------------- Error Message
--------------------------------------------------------------
[0]PETSC ERROR: Object is in wrong state
[0]PETSC ERROR: Not for unassembled matrix
[0]PETSC ERROR: See https://www.mcs.anl.gov/petsc/documentation/faq.html
for trouble shooting.
[0]PETSC ERROR: Petsc Release Version 3.13.1, unknown
[0]PETSC ERROR: ./transpose-petsc on a  named jrhammon-nuc by jrhammon Mon
Jun  1 17:01:05 2020
[0]PETSC ERROR: Configure options --prefix=/home/jrhammon/PRK/deps/petsc
--with-blaslapack-dir=/opt/intel/inteloneapi/mkl/2021.1-beta06
--with-mpi-dir=/opt/intel/inteloneapi/mpi/2021.1-beta06 --with-cxx=0
--with-fc=0
[0]PETSC ERROR: #1 MatTranspose() line 4878 in
/home/jrhammon/PRK/deps/petsc-src/src/mat/interface/matrix.c
[0]PETSC ERROR: #2 main() line 169 in transpose-petsc.c
[0]PETSC ERROR: PETSc Option Table entries:
[0]PETSC ERROR: -i 10
[0]PETSC ERROR: -n 100
[0]PETSC ERROR: ----------------End of Error Message -------send entire
error message to petsc-maint at mcs.anl.gov----------
Abort(169073) on node 0 (rank 0 in comm 16): application called
MPI_Abort(MPI_COMM_SELF, 169073) - process 0

Jeff

On Mon, Jun 1, 2020 at 4:11 PM Jed Brown <jed at jedbrown.org> wrote:

> I would just use MatTranspose here.  We've had limited demand for more
> sophisticated operations on objects of type MATTRANSPOSE and there
> probably isn't much benefit in fusing the parallel version here anyway.
>
> Jeff Hammond <jeff.science at gmail.com> writes:
>
> > I am trying to understand how to use a transposed matrix view along the
> > lines of Numpy (
> >
> https://github.com/ParRes/Kernels/blob/master/PYTHON/transpose-numpy.py#L99
> > ).
> >
> >
> https://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatCreateTranspose.html
> > says
> >
> > "Creates a new matrix object that behaves like A'"
> >
> > But later it seems to suggest that this is only true when performance a
> > matrix-vector product:
> >
> > "Rather the new matrix object performs the matrix-vector product by using
> > the MatMultTranspose() on the original matrix"
> >
> > I got the idea to use this view concept from
> >
> https://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatTranspose.html
> ,
> > which says:
> >
> > "Consider using MatCreateTranspose() instead if you only need a matrix
> that
> > behaves like the transpose, but don't need the storage to be changed."
> >
> > When I try to use the transpose view in MatAXPY using this code:
> >     Mat AT;
> >     ierr = MatCreateTranspose(A, &AT); // create transpose view of A
> >     ierr = MatAXPY(B, one, AT, SAME_NONZERO_PATTERN);
> >
> > I get an error suggesting that I am doing something bad:
> >
> > [0]PETSC ERROR: --------------------- Error Message
> > --------------------------------------------------------------
> > [0]PETSC ERROR: Object is in wrong state
> > [0]PETSC ERROR: Not for unassembled matrix (Y)
> > [0]PETSC ERROR: See https://www.mcs.anl.gov/petsc/documentation/faq.html
> > for trouble shooting.
> > [0]PETSC ERROR: Petsc Release Version 3.13.1, unknown
> > [0]PETSC ERROR: ./transpose-petsc on a  named jrhammon-nuc by jrhammon
> Mon
> > Jun  1 15:38:20 2020
> > [0]PETSC ERROR: Configure options --prefix=/home/jrhammon/PRK/deps/petsc
> > --with-blaslapack-dir=/opt/intel/inteloneapi/mkl/2021.1-beta06
> > --with-mpi-dir=/opt/intel/inteloneapi/mpi/2021.1-beta06 --with-cxx=0
> > --with-fc=0
> > [0]PETSC ERROR: #1 MatAXPY() line 74 in
> > /home/jrhammon/PRK/deps/petsc-src/src/mat/utils/axpy.c
> > [0]PETSC ERROR: #2 main() line 170 in transpose-petsc.c
> > [0]PETSC ERROR: PETSc Option Table entries:
> > [0]PETSC ERROR: -i 10
> > [0]PETSC ERROR: -n 100
> > [0]PETSC ERROR: ----------------End of Error Message -------send entire
> > error message to petsc-maint at mcs.anl.gov----------
> >
> > What do I need to do to use a transpose view properly outside of M*V?
> >
> > Thanks,
> >
> > Jeff
> >
> > --
> > Jeff Hammond
> > jeff.science at gmail.com
> > http://jeffhammond.github.io/
>


-- 
Jeff Hammond
jeff.science at gmail.com
http://jeffhammond.github.io/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20200601/edf185c4/attachment.html>