Solution difference of ksp linear solver by using "singleton" MPI and mpirun -np 1 to run ogs
Description
When PETSc ksp linear solver and pc are used for sequential computation, the linear solver under running the code as "singleton" MPI (non mpirun) and by command mpirun -np 1
, respectively, has different convergence behavour for some complicated ksp+pc settings. Also see the discussions in MR !4012 (merged).
So far, it is not clear which part leads to that problem or using mpirun -np 1
is mandatory for PETSc.
Edited by wenqing