[petsc-dev] TS Terminology

Matthew Knepley knepley at gmail.com
Fri Oct 20 14:16:35 CDT 2017


On Fri, Oct 20, 2017 at 1:20 PM, Jed Brown <jed at jedbrown.org> wrote:

> Barry Smith <bsmith at mcs.anl.gov> writes:
>
> >> On Oct 20, 2017, at 11:53 AM, Jed Brown <jed at jedbrown.org> wrote:
> >>
> >> Barry Smith <bsmith at mcs.anl.gov> writes:
> >>
> >>>   The name absolutely has to be changed. But to what? And the manual
> page is WRONG! You cannot justify that no matter how much you want to keep
> the current confusing/inaccurate name.
> >>
> >> Are you also adamant that SNESComputeFunction must be changed?  After
> >> all, it isn't returning the output of the function that was passed to
> >> SNESSetFunction.  If SNESComputeFunction is okay, but TSComputeIFunction
> >> is not, what is the rationale for that?
> >
> >   In TS it is DAMN!!!! confusing. (Since you and Emil have lived with it
> from day one I know it is not confusing to you; but it is confusing to
> everyone else).
>
> Why is it DAMN!!!! confusing in TS, but exactly the same pattern is
> "very minor" in SNES?  We need an explanation here unless we're going to
> rename all XComputeY() functions to compute exactly Y.
>
> Is it possible it has something to do with you having spent 30 years
> thinking deeply about Newton solvers and this moving the RHS vector over
> just seems like a trivial and obvious transformation?


I think you are right that the SNES thing needs fixed as well. The problem
has not
come up because no one ever uses that interface. It is only there because
it makes
FAS a lot easier.

   Matt


> >   In SNES it is a very minor confusion.
> >
> >   We absolutely need to fix things that are DAMN!!!! confusing. Fixing
> things that are minor confusing is much less important. So it would be fine
> to change SNESComputeFunction() but I have no reason to be adamant about it.
> >
> >   Barry
>



-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener

https://www.cse.buffalo.edu/~knepley/ <http://www.caam.rice.edu/~mk51/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20171020/410d8609/attachment.html>


More information about the petsc-dev mailing list