Matthew Knepley <[email protected]> writes:

> When we call PCSetUp() on a FieldSplit, it does not call SetUp() on its
> subsolves. Is this deliberate? It is making it hard to test a MF MG on the
> A11 of a fieldsplit.

I don't know what was intended, but it's a problem to set everything up
when the inner solves will fail (e.g., bad pivot) without getting them
out to set other options.

Attachment: pgpZf6o6XovaP.pgp
Description: PGP signature

Reply via email to