testing limiting wqtotal for MultisegmentWellPrimaryVariables #6701
Conversation
GitPaean
added
the
manual:irrelevant
|
jenkins build this failure_report please |
GitPaean
force-pushed
the
testing_limiting_wqtotal
branch
from
ff10ccb to
4af9514
Compare
|
jenkins build this failure_report please |
|
the regression results do not look worrying (most likely) |
GitPaean
force-pushed
the
testing_limiting_wqtotal
branch
from
4af9514 to
2f5269c
Compare
| // value_[seg][WQTotal] = old_primary_variables[seg][WQTotal] - relaxation_factor * dwells[seg][WQTotal]; | ||
| auto dx = relaxation_factor * dwells[seg][WQTotal]; | ||
| const Scalar sign = dx > 0.? 1. : -1.; | ||
| dx = sign * std::max(std::min(1., std::abs(dx)), 0.4 * std::abs(dx)); |
There was a problem hiding this comment.
This will fail to compile if Scalar is different from double (the std::min() and std::max() function templates cannot be specialised when applied to objects of different types).
That said, it's not immediately clear to me what this is trying to do. Are you trying to leave "small" dx values unchanged and limit "large" values? If so, I'd probably write it
as
if (std::abs(dx) > some_threshold) {
dx = sign * std::max(some_threshold, static_cast<Scalar>(0.4) * std::abs(dx));
}instead.
There was a problem hiding this comment.
Thanks for the comments regarding the Scalar issue. It is not in any way the PR will go in with this types of design. It is mostly for discussion and communication with the people working together. And also 1. is a relatively big value (it is per second), but it was too small for the cases we are looking at during the Newton iterations.
2 participants
No description provided.