Problem with with convert for inplace conversion

[Benoit NENNIG]

Hi, I observe that the Mat.convert method does not behave as I would expected for inplace conversion.

#Create the Nested Matrix
C=PETSc.Mat().createNest([[A, A], [None, 2*A]])
C.convert(mat_type=PETSc.Mat.Type.AIJ)  # fails

It seems not possible to rewrite the matrix inplace (except with mpirun -n 1). However the synthax

B=PETSc.Mat()
C.convert(PETSc.Mat.Type.AIJ,B) 

with an auxiliary B matrix works fine.

Sorry if it is due to a misconception, but inplace conversion work fine with transpose and both methods seem to share the same interface with 'reuse' and MAT_INPLACE_MATRIX

C.transpose() # Ok!

info : petsc4py, Version: 3.9.0, arch-linux2-c-optim, complex

Benoit

[Lisandro Dalcin]

This is not a petsc4py issue. It seems that MatConvert() for MATNEST with reuse=MAT_INPLACE_MATRIX is not supported in core PETSc.

@BarryFSmith @jedbrown @stefano_zampini What should we do about it?

[Jed Brown]

It needs to be implemented. I think it's pretty easy, but I don't have any free cycles for at least a week or two.

[hadjout said aissa]

is there a possibility for this bug to be fixed ?

[Lisandro Dalcin]

This is not a petsc4py issue. You should submit a bug report/feature request to PETSc https://gitlab.com/petsc/petsc/-/issues .

[Lisandro Dalcin]

Submit bug reports and feature requests to https://gitlab.com/petsc/petsc/-/issues