Initialize old solution with the current solution on restart when it is not found in checkpoint #30135
Conversation
|
@YaqiWang I think I understand restarting systems a bit better now, and I feel like we can easily implement a flexible restart where we can specify variables we want to restart. It's currently looping over all the variables found in the header |
|
Job Documentation, step Docs: sync website on 875842d wanted to post the following: View the site here This comment will be updated on new commits. |
…is not found in checkpoint idaholab#30111
| @@ -19,6 +19,9 @@ | |||
| const std::string RestartableEquationSystems::SystemHeader::system_solution_name = | |||
| "SYSTEM_SOLUTION"; | |||
|
|
|||
| const std::string RestartableEquationSystems::SystemHeader::system_old_solution_name = | |||
There was a problem hiding this comment.
Is there a static definition of this vector name somewhere else? It would be nice to not hard code it here if we can get it from somewhere else
There was a problem hiding this comment.
I see Moose::OLD_SOLUTION_TAG
There is a failing Bison test |
We want to be able to just load variables in a checkpoint file into individual variables like for the weighting functions in IQS with adjoint solution. |
| @@ -264,6 +267,9 @@ RestartableEquationSystems::load(std::istream & stream) | |||
|
|
|||
| bool modified_sys = false; | |||
|
|
|||
| bool init_old_solution = !sys_header.vectors.count(SystemHeader::system_old_solution_name) && | |||
There was a problem hiding this comment.
We will need to do the same for older.
There was a problem hiding this comment.
If checkpoint only has current and old, which should go into older?
There was a problem hiding this comment.
That can be an error? Otherwise, I would say old go to older.
There was a problem hiding this comment.
If checkpoint only has current, current goes to older? That sounds too many conditionals.
|
Does the restart happen late enough that all objects have been initialized that request old and older states? |
Time integrator requests this in its constructor. I guess it will be safer if we throw an error if an object tries to do this when |
|
@loganharbour do you want to dig into this? Otherwise I can |
|
@YaqiWang and @NamjaeChoi have you had more time to think about this? |
|
@loganharbour I think Namjae and me are waiting for a help. |
Closes #30111.