[petsc-dev] TS confusion

[Barry Smith]

  Thanks

On Sep 15, 2013, at 10:19 AM, Jed Brown <jedbrown at mcs.anl.gov> wrote:

> Barry Smith <bsmith at mcs.anl.gov> writes:
> 
>>   I am ONLY providing RHSJacobian and RHSFunction and not IJacobian at all. I am using ARKIMEX but have called 
>>  ierr = TSSetType(ts,TSARKIMEX);CHKERRQ(ierr);
>>  ierr = TSARKIMEXSetFullyImplicit(ts,PETSC_TRUE);CHKERRQ(ierr);
> 
> This logic just wasn't optimized.  I was thinking that
> TSARKIMEXSetFullyImplicit implied that the user had set both RHSJacobian
> and IJacobian.  So it should be safe to update that logic to only do the
> duplication if both are provided.
> 
>>   You can run
>>   src/ts/examples/tutorials/advection-diffusion-reaction/ex10.c and
>>   put a breakpoint in it and you will see it is calling the MatAXPY()
>>   as I indicated. I suspect the logic of fully implicit TSARKIMEX is
>>   far from optimized when only RHSJacobian is provided.
>> 
>>    What TSType should I use?
> 
> What method do you want?  I would use ARKIMEX and just optimize that
> logic.  (I'm about to drop offline until evening, but I'll get to it if
> you don't.)