benchcab demo/training (Land surface Modelling Community Working Group Workshop)

[SeanBryan51]

benchcab demo

Prerequisites:

You will need:

1. An NCI account with access to projects hh5 and ks32
2. A modelevaluation.org account
3. A GitHub account (this is not strictly necessary but is recommended)

Running a quick and easy test suite:

benchcab is implemented as a command line tool and is available in the hh5 conda environment. To activate the environment, run:

module use /g/data/hh5/public/modules
module load conda/analysis3-unstable

To run the test suite, benchcab must be run from an example work directory under either /scratch or /g/data.

Change directory into /scratch/<project>/<user>/:

cd /scratch/<project>/<user>/

Clone the example work directory:

git clone git@github.com:CABLE-LSM/bench_example.git

Next, edit the config.yaml configuration file in the example work directory. Here is where we configure the test suite for benchcab:

project: w97

experiment: five-site-test

realisations: [
  {
    path: "trunk",
  },
  {
    path: "branches/Users/ccc561/v3.0-YP-changes",
  }
]

modules: [
  intel-compiler/2021.1.1,
  netcdf/4.7.4,
  openmpi/4.1.0
]

• The project key specifies the NCI project to charge the simulations too.
• The experiment key specifies the experiment we want to run. Experiments are defined in the benchcab-evaluation workspace on modelevaluation.org. The experiment dictates which site datasets we use to drive our models in our test suite.
• The realisations key specifies a list of branches we want to test. One or more branches can be specified in the list.
• The modules key specifies which modules to use for compiling CABLE executables. There are also a number of optional parameters in the configuration file. For more details, see the documentation on config.yaml options.

In this example we will run a simple test suite for the AU-Tum Tumbarumba site. Set the project key to your project of choice and set experiment to AU-Tum.

To run the test suite, run the following command from the example work directory:

benchcab run --verbose