[petsc-dev] major changes to DA in PETSc-dev

[Jed Brown]

For the sake of argument, why not get rid of all static types that inherit
from PetscObject?  Then everything in Mat, KSP, SNES, etc could use
QueryFunction for dispatch.

My claim is that the relationship between DM and it's subtypes, from the
user's perspective, is much more like the relationship between PetscObject
and it's subtypes (e.g. Mat).  There are some functions that operate on
generic PetscObjects (e.g. PetscOList*, View, SetFromOptions), but most uses
of Mat need to know statically that it is a Mat (not a KSP or some other
PetscObject) in the same way that most uses (by users) of DA need to know
statically that it is a DA and not some other DM.

This is why I prefer the more static inheritance model from PetscObject ->
Mat for this purpose.  What flexibility is being gained by using this new
dynamic model?

Jed
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20101018/1db59f83/attachment.html>