[petsc-dev] Revamping the multilevel KSP interface

[Barry Smith]

On Mar 17, 2012, at 10:41 PM, Jed Brown wrote:

> On Sat, Mar 17, 2012 at 21:34, Jed Brown <jedbrown at mcs.anl.gov> wrote:
> Is grid sequencing the only use case? What is the plan with grid sequencing for linear problems? Are the people who really want to do that capable of using SNES with KSPONLY? That would also make it easier to use a higher order prolongation the first time a given grid is reached, but use standard interpolants in subsequent cycles (as Achi does for the best work estimates).
> 
> And does anyone prefer this interface where they set a callback to build the RHS instead of calling the function themselves (or constructing the RHS any way they please)? I know it makes for a somewhat slick example, but the user has to interact with a Vec anyway to get the solution out.
> 
> http://www.mcs.anl.gov/petsc/petsc-dev/src/ksp/ksp/examples/tutorials/ex25.c.html

  Jed,

  As I noted before to you I did this to have a similar interface for KSP as SNES so that people could see the pattern and easily "switch up" to SNES from KSP. 

  I know you advocate having the user just use SNES if they want this cool "SNES-like" behavior. 

  Will it kill you to leave this stuff in there? Are you wanting to remove it just for philosophical reasons or is it causing trouble to have it?

   Barry