PHOLD is a well known PDES benchmark. This repository contains two versions:
- pholdio: which implements the ross-org/RIO API to demonstrate proper use of the checkpointing functionality
- dpholdio: a version of phold-delta which implements the RIO API. LPs in this model contain a large state.
The dpholdio model can be built with or without the RIO library (the USE_RIO option in CMake).
The pholdio model requires the RIO library.
Both models operate with the same command line flags as PHOLD.
There is one additional flag: --io-store=[0,1].
When --io-store=0, the models will attempt to start from a checkpoint named "pholdio_checkpoint". This is the default value for io-store.
When --io-store=1, the models will save a checkpoint with the name "pholdio_checkpoint" after the simulation has completed.