[petsc-users] is TS_EQ_DAE_SEMI_EXPLICIT_INDEX functional

[Constantinescu, Emil M.]

Indeed, various time steppers can take advantage of the differential 
form provided and also can serve as a sanity check (e.g., warn users 
before they use an explicit solver on an index-2 DAE). To my knowledge, 
we do not have solvers that take advantage of semi-explicit DAEs, but 
it's good practice to annotate applications for when solvers are 
available. In ARKIMEX, we check only if the problem is TS_EQ_IMPLICIT or 
not (this includes DAEs and ODEs with mass matrix on the left M\dot{u}). 
If you are solving a DAE, you should use TS_EQ_IMPLICIT. G=M^-1 g is 
either for nonstiff ODE or partitioned ODE with mass matrix, which can 
be used if you have a fast way to compute M^-1 g.

Emil

On 9/2/19 4:44 PM, Jed Brown wrote:
> I believe this is intended to work with most any implicit solver,
> *provided* the initial conditions are compatible.  It was added by Emil,
> but I don't see it explicitly tested in PETSc.
>
> "Huck, Moritz via petsc-users" <petsc-users at mcs.anl.gov> writes:
>
>> Hi,
>> TS_EQ_DAE_SEMI_EXPLICIT_INDEX(?) are defined in TSEquationType but not mentioned in the manual.
>> Is this feature functional ?
>> If yes how do I have to define the RHSFunction?
>> (I am asking since the ODE variant has it defined as G= M^-1 g, which cannot work for a DAE)
>>
>> Best Regards,
>> Moritz