Question: is it possible to run OpInf-FOM-OpInf coupling?

[ikalash]

This is a question for @eparish1 : is there a plan to make it possible to run an OpInf-FOM-OpInf coupling when a 3 subdomain DD is employed? When I try this for the laser weld problem, I get the following error:

[ikalash@hayka M20_reg1em8_OpInf-FOM-OpInf]$   julia --project=@. /home/ikalash/Norma.jl-Eric/src/Norma.jl  laser-weld.yaml | tee out
[ Info: Debugging disabled
Reading simulation file: laser-weld.yaml
Reading subsimulation file: holder-0.yaml
Reading subsimulation file: gauge.yaml
ERROR: LoadError: Multidomain subdomains must all have the same physics
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:35
 [2] Main.Norma.MultiDomainSimulation(params::Dict{String, Any})
   @ Main.Norma ~/Norma.jl-Eric/src/simulation.jl:122
 [3] create_simulation(params::Dict{String, Any}, name::String)
   @ Main.Norma ~/Norma.jl-Eric/src/simulation.jl:23
 [4] create_simulation(input_file::String)
   @ Main.Norma ~/Norma.jl-Eric/src/simulation.jl:34
 [5] run
   @ ~/Norma.jl-Eric/src/Norma.jl:33 [inlined]
 [6] top-level scope
   @ ~/Norma.jl-Eric/src/Norma.jl:45
in expression starting at /home/ikalash/Norma.jl-Eric/src/Norma.jl:7

[eparish1]

@ikalash I'll diagnose and get this fixed.

[ikalash]

Thank you @eparish1 ! It might be that just a throw needs to be removed. I would think you can do OpInf-FOM-OpInf if the requirement is that an OpInf model be coupled always to a FOM, not another OpInf.

[eparish1]

Yes, you might just be able to drop the throw. I remember not really liking this bit of the code though, so I'll take a look and see if I can come up with a better solution.

[ikalash]

@eparish1 : sounds good. I haven't looked at the relevant parts of the code yet, so dropping the throw may not be sufficient.

[eparish1]

@ikalash @lxmota

The issue is just because of the throws, and it can be mitigated by commenting these out. However, I think it raises the bigger issue of how we want to handle this. The "best" way is to probably add another field to the OpInf ROM corresponding to its FOM model type. The only negative thing about this is its another field the user needs to put in. The easiest thing is to just remove the throws for only allowing Schwarz if the models are the same type. It looks like the code just quits if they aren't the same type anyway. I'm happy to go with whatever path. Any preferences?

[lxmota]

I removed all that code, see sandialabs/Norma.jl@967824e76e334357f72d25c745905672537878f0 for reasons.

[ikalash]

Thanks for making the change @lxmota . I will try again the OpInf-FOM-OpInf coupling soon to verify that it works now.