Fix handling for default algorithms for DualLinearProblems #775
Conversation
| function SciMLBase.init(prob::DualAbstractLinearProblem, alg::DefaultLinearSolver, args...; kwargs...) | ||
| if alg.alg === DefaultAlgorithmChoice.GenericLUFactorization | ||
| return __init(prob, alg, args...; kwargs...) | ||
| else | ||
| return __dual_init(prob, alg, args...; kwargs...) | ||
| end |
There was a problem hiding this comment.
Tests added. Looks good on 1.11 but fails on LTS
|
tests fail |
|
CI is having some issues here, some are complaining about Pkg not being available, and one says that Julia isn't a recognized executable. |
|
NoPre is a real failure |
|
So the normal default algorithm But the default algorithm |
jClugstor
force-pushed
the
dual_genericLU_bypass_default
branch
from
c515eb3 to
84b2a6b
Compare
jClugstor
force-pushed
the
dual_genericLU_bypass_default
branch
from
84b2a6b to
eab1f06
Compare
jClugstor
force-pushed
the
dual_genericLU_bypass_default
branch
from
eab1f06 to
26f36ce
Compare
|
@ChrisRackauckas I'm not sure what to do here, the JET tests pass for 1.11, but not LTS, so whether I mark the test as broken or not there will still be a test failure. |
|
can you mark it broken with a |
|
Yeah that works if that's ok to do. |
|
@ChrisRackauckas I think this is now good to go. |
|
@hersle now try that benchmark. |
3 participants
Checklist
contributor guidelines, in particular the SciML Style Guide and
COLPRAC.
Additional context
This makes it so that the default alg is chosen based on the primal A and b before going to make the Dual cache. This also means that for small problems GenericLU will be used, and the cache will be a normal LinearCache.