[CIG-SHORT] Pylith turn off vtk writers
Brad Aagaard
baagaard at usgs.gov
Thu Aug 23 08:37:25 PDT 2018
On 08/23/2018 08:29 AM, Francisco Delgado wrote:
> Another question, why should I use*ksp_rtol=ksp_atol=1.0e-12* instead
> of 1e-11 or 1e-10? Again, this can save a lot of time
You can relax your solver tolerances. You want to make sure the solution
is accurate relative to whatever observations you are comparing against.
Remember the solver tolerances are apply to the nondimensionalized problem.
You should also check to make sure you are using a good preconditioner.
I recommend reading the discussion of the PETSc solver settings in the
PyLith manual, especially for the 2D and 3D subduction zone examples
where we try to use the best preconditioners.
If your simulation uses a fault with friction, then you need to be extra
careful. In order for the fault to "stick" you need to make sure the
fault zero_tolerance is greater than the absolute tolerance in the
linear solve. See the online PyLith tutorials on friction for more
information.
Regards,
Brad
More information about the CIG-SHORT
mailing list