[petsc-dev] Question about the fully implicit Runge-Kutta methods (TS)

[Jed Brown]

Matthew Knepley <knepley at gmail.com> writes:

> On Tue, Oct 5, 2021 at 12:42 AM Wang Jingyuan <mc05569 at connect.um.edu.mo>
> wrote:
>
>> Dear Sir or Madam,
>>
>> I am a student of the University of Macau, I am trying to learn PETSc and
>> I have two questions about the TSIRK.
>>
>> 1.PETSc 3.16.0 adds “irk" for the TSType and it offers many options like TSIRK1BEE,
>> TSIRK2C, TSIRK2D, TSIRK2E, TSIRK3, TSIRKL2,
>>
>
> The types you mention above do not exist. For instance, the first one
> "TSIRK1BEE" is actually
> https://petsc.org/main/docs/manualpages/TS/TSARKIMEX1BEE.html

As Ed Bueler has commented, these non-existent types are incorrectly listed in the man page (people like me missed those copy/paste remnants during code review).

https://petsc.org/main/docs/manualpages/TS/TSIRK.html

The 1BEE and related types are diagonally implicit, so don't make sense under TSIRK. We would like to add Radau and Lobatto methods to PETSc -- it would be a relatively easy contribution.