use OrdinaryDiffEq's DAE initialization. #401
Conversation
|
Add a test? I don't see how this works, and it doesn't disable the IDA initialization. |
|
yeah, I wasn't sure what to do about the IDA initialization. The options I see are to add it as it's own initialization that can be used by anyone, stop using it entirely, or something else. Thoughts? |
|
This is now much closer to working. Tests still seem to be failing though. Something is slightly off. |
|
Rebase and add tests |
|
This is almost ready. The remaining problem is that the OrdinaryDIffEq |
|
The test failure looks real |
Codecov Report
@@ Coverage Diff @@
## master #401 +/- ##
==========================================
+ Coverage 80.21% 81.31% +1.09%
==========================================
Files 11 11
Lines 1466 1472 +6
==========================================
+ Hits 1176 1197 +21
+ Misses 290 275 -15
... and 2 files with indirect coverage changes 📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
|
test added. |
| function handle_callback_modifiers!(integrator::IDAIntegrator) | ||
| IDAReInit(integrator.mem, integrator.t, integrator.u, integrator.du) | ||
| DiffEqBase.initialize_dae!(integrator) | ||
| DiffEqBase.initialize_dae!(integrator, IDADefaultInit()) |
There was a problem hiding this comment.
I thought that after a callback the default initialization was the one we wanted. Is it not?
There was a problem hiding this comment.
I don't recall if it was using the user-chosen one or the default.
There was a problem hiding this comment.
OrdinaryDIffEq just calls integrator.reeval_fsal = true here. Previously, Sundials would just call DiffEqBase.initialize_dae!(integrator) which is now spelled DiffEqBase.initialize_dae!(integrator, IDADefaultInit()) (since now the initialization is customizable.
|
Can we merge this? |
|
It's missing the lower bound bump. |
|
version bumped. |
|
It looks like your test case fails? |
Sundials.jl got an implementation of the OrdinaryDiffEq DAE initialization interface in SciML#401. Unfortunately, it turns out that this doesn't quite work, because the u/du in Sundials' integrator struct are not the actual memory that IDA uses (it has an internal copy). In [1], I propose re-using `u_modified` to inform us that the shadow copy was modified during initialization. This is the Sundials side of that interface. [1] SciML/OrdinaryDiffEq.jl#1969
Sundials.jl got an implementation of the OrdinaryDiffEq DAE initialization interface in SciML#401. Unfortunately, it turns out that this doesn't quite work, because the u/du in Sundials' integrator struct are not the actual memory that IDA uses (it has an internal copy). In [1], I propose re-using `u_modified` to inform us that the shadow copy was modified during initialization. This is the Sundials side of that interface. [1] SciML/OrdinaryDiffEq.jl#1969
Sundials.jl got an implementation of the OrdinaryDiffEq DAE initialization interface in SciML#401. Unfortunately, it turns out that this doesn't quite work, because the u/du in Sundials' integrator struct are not the actual memory that IDA uses (it has an internal copy). In [1], I propose re-using `u_modified` to inform us that the shadow copy was modified during initialization. This is the Sundials side of that interface. [1] SciML/OrdinaryDiffEq.jl#1969
Sundials.jl got an implementation of the OrdinaryDiffEq DAE initialization interface in #401. Unfortunately, it turns out that this doesn't quite work, because the u/du in Sundials' integrator struct are not the actual memory that IDA uses (it has an internal copy). In [1], I propose re-using `u_modified` to inform us that the shadow copy was modified during initialization. This is the Sundials side of that interface. [1] SciML/OrdinaryDiffEq.jl#1969
3 participants
This allows for custom initialization functions when using
IDA.