Skip to content

MPI and noMPI inversion lead to different results #18

Description

@leobpujol

Cost function calculation appears to work as intended (calc_innovation).

Computation of some dof%h values using the first modified control vector, i.e. resulting from the initial gradient, are not consistent with a direct simulation using the same parameters (but seem to be hydraulically plausible).

  • Modified parameters, from the control vector, seem to be passed on to the direct model correctly for the second direct run of the iterative process. Then why are the results from the solver inconsistent ?

-> Issues with other structures handled by MPI ? Like BCs ? Initial states ?

Tested with 4 procs, on Garonne Marmandaise case, using WSE observations with i) 10 porosity patches, ii) 1 manning patch
Tested with 2 procs ... with ii) 1 manning patch : assim seems to stall.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions