Adding option to swap Neumann and Dirichlet BCs for Schwarz non-overlapping coupling and contact #39
Conversation
non-overlapping Schwarz iteration with Dirichlet or Neumann BC. Also adding option to swap the BCs on the fly. Please see examples/nonoverlap/static-same-time-step for an example of the syntax. Also switched this problem to start from Neumann BC instead of Dirichlet.
There was a problem hiding this comment.
Looks good to me and request minor changes: ics_bcs.jl:189 change "sting" to "string" and delete lines 364 and 367. Thanks for checking that this swapping also affects contact negatively. Not sure why Daria reported that there was no change, but she used a much earlier version of the code.
I just addressed these. Thanks for catching the typo. Not sure about why the behavior was different before. Now the method maxes out at 16 iterations, the max specified in the input file. It is a bit disconcerting but I suppose the swapping is not really a standard thing with a known result on how it will affect performance. |
|
Sure! Merging now. |
Description in title. The swapping hurts in terms of # of Schwarz iterations for both contact and coupling. I did not add a test that executes the new option, as it is not competitive, though examples/nonoverlap/static-same-time-step shows the syntax. I can add a test if this is desired.