Closed ma-sadeghi closed 1 year ago
It's quite hard to calculate these equivalences, however, we decreased the tolerance on taufactor 2 until the converged value was closer to the 'true' value (calculated with a very low tolerance) than any of the other solvers. Let us know your thoughts on whether this is sufficient.
In that case, then your actual speedups should be slightly better (when done using an equivalent flux tolerance), which addresses my comment. Thanks!
IIRC, TauFactor uses
5e-2
for flux tolerance. Have you made sure that you use an equivalent tolerance for the packages used in the speed comparison chart? I don't know about the rest, but I think PoreSpy uses1e-8
for "residual" tolerance (might need some trial and error to see how that translates to flux tolerance).