You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In trying to figure out why #5291 does not test cleanly, I've come across an issue in the Simulator<dim>::assemble_and_solve_stokes() function. Here is how it looks, in redux on the current main branch:
The bug (?) is that the computation of the nonlinear residual (and putting it into *initial_nonlinear_residual only happens if compute_initial_residual==true. But later on, we use that value.
There are two ways to see this issue:
Either it's a bug where we use a value we haven't computed.
Or it's on purpose -- I'm leaning towards that. The idea then is that compute_initial_residual is only true in the very first iteration, and in all other iterations we scale the computed residual by the value computed the very first time around. I think that's what the variable name represents.
I'm going to fix up #5291 based on that second assumption by separating the initial_nonlinear_residual into two variables: One we're asking to compute, and one that is passed in by value.
In trying to figure out why #5291 does not test cleanly, I've come across an issue in the
Simulator<dim>::assemble_and_solve_stokes()
function. Here is how it looks, in redux on the currentmain
branch:The bug (?) is that the computation of the nonlinear residual (and putting it into
*initial_nonlinear_residual
only happens ifcompute_initial_residual==true
. But later on, we use that value.There are two ways to see this issue:
compute_initial_residual
is onlytrue
in the very first iteration, and in all other iterations we scale the computed residual by the value computed the very first time around. I think that's what the variable name represents.I'm going to fix up #5291 based on that second assumption by separating the
initial_nonlinear_residual
into two variables: One we're asking to compute, and one that is passed in by value.@MFraters Does that match your interpretation?
The text was updated successfully, but these errors were encountered: