endOfTransportStep output policy leads to desynchronization of models if executed in parallel
Summary
When selecting transport.output.policy = endOfTransportStep, and running the application in parallel, the models desynchronize. The transport model may then take time steps which advance further than the Richards model.
Steps to reproduce
- Model settings:
simulation.mode = richards+transport,transport.output.policy = endOfTransportStep - Run in parallel:
dorie run -p 2 config.ini
What is the current bug behaviour?
Transport model advances further in time than Richards model if it outputs every time step and the simulation is run in parallel.
What is the expected correct behaviour?
Transport model never advances further in time than Richards model.
Relevant logs, screenshots, files...?
Will be supplied later
Reproducing input
| Input data | |
|---|---|
| Simulation Case | Description goes here |
| PFG config file | if any |
| Grid mapping file | if any |
| GMSH grid file | if any |
| Boundary Condition file | |
| Parameterization file | |
| Run config file |
Ideas how to fix this?
Check how write_data() method or endOfTransportStep output policy of transport model mess with time.
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information