OpenMDAO / dymos / 6087501337

91%

Build Type

push

github

Committed by web-flow

Commit Message

Fixed some issues that were causing simulation to fail under MPI. (#978)

* working on simulation under mpi. works with linked phases but connections problematic.

* Fixed simulation under MPI.

simulation was failing under MPI.
- Setting simulation phase values from their source phases now accounts for MPI.
- This code is built on the assumption that a phase and its corresponding simulation phase are on the same processor.
- The trajectory created by `Trajectory.simulate` will not bother setting up linkages or interphase connections since they are irrelevant for simulation.
- Bumped the minimum OpenMDAO version as specified in setup.py to 3.26.  We were already using this as the oldest supported version on CI.
- In `set_vals_from_phase` for simulation phase we now don't grab variable values from their ultimate source (from_src=False).
- The finite burn MPI test cases now include simulation.

* Pass trajectory's comm to the subproblem used for simulation.

* cleanup

* Added unit simplification to introspection when units involve multiplication by time units.
This prevents units such as '1/s*s' intead of None.

Added a warnings context to simulation setup to ignore setup and unused options warnings that are not the fault of the user.

* Cleaned up simulation phase so that state values can be pulled from either states or initial_states in the phase being simulated.

* If pulling initial states from initial states, do not do any slicing

Run Details

50 of 53 new or added lines in 6 files covered. (94.34%)

28 existing lines in 1 file now uncovered.

32710 of 35310 relevant lines covered (92.64%)

4.36 hits per line