Barry Smith <[email protected]> writes:

>   blocks are always square. 
>
>   If the user wants non-square blocks then they go though the long involved 
> construction process

non-square blocks show up in shell matrices for interpolation and
restriction, which is not exactly a weird application.  I still think
block size is an extraneous concept to most users, thus I don't think
that adding it to all the convenience wrappers is a welcome or
clearly-beneficial change.

What problems would we be inviting by making MatGetBlockSize return
PetscAbs(rmap->bs)?  Obviously there are a number of places where we
access the field directly, but those should use the interface anyway.

Attachment: pgp9eeap1niJC.pgp
Description: PGP signature

Reply via email to