Up to now, I've not changed any API
http://petsc.cs.iit.edu/petsc/temp/petsc-dev-ts/rev/86e0cf3edea1
The Algorithm pattern appears (Barry, let me use such "pretension
academic language"), so I think we should not duplicate the code with
the time-stepping loop, and use something like this:
TSSolve(x):
TSSetUp()
#This subtype handles the whole time-stepping loop (like Sundials?)*/
if ts->ops->solve {
ts->ops->solve()
return
#This subtype handles step by step (many PETSc TS subtype, like Theta) */
for (i=0; i< max_steps; i++)
if i==0: Monitor()
PreStep()
ts->ops->step( .... ) #
PostStep()
Monitor()
Moreover, we could redefine TSStep() to mean "do just one (or many)? steps":
TSStep(ts, step, ....)
if step ==0: Monitor()
PreStep()
ts->ops->step( .... ) #
PostStep()
Monitor()
and reuse this for TSSolve()... TSStep() would then give users a bit
more of control, letting them to manage the loop outside TSSolve(),
but still in a high-level way...
Coments ?
--
Lisandro Dalcin
---------------
CIMEC (INTEC/CONICET-UNL)
Predio CONICET-Santa Fe
Colectora RN 168 Km 472, Paraje El Pozo
Tel: +54-342-4511594 (ext 1011)
Tel/Fax: +54-342-4511169
