Resolve "Use BCGS_AMG_SSOR solver for finite volume methods"
What does this MR do?
Switch to the BCGS_AMG_SSOR linear solver backend for finite volume computations.
How does this affect runtime?
There is no observable change in the system test runtime. I ran a single 2D simulation on a heterogeneous domain and observed a runtime of around 185s vs the original 202s. This is not a real benchmark, but it indicates that we can have a speedup of 0 to 10%. I think it's therefore correct to switch to the new backend.
Is there something that needs to be double checked?
Can this MR be accepted?
- Implemented BCGS_AMG_SSOR linear solver backend
- Pipeline passing
- Squash option set
- Delete branch option set
Added entry to
Assignee: If the Squash option is set, check/update the commit message right before merging!
Closes #169 (closed)