Launching the benchmarks

The feelpp-benchmarking-exec command is used to launch the benchmarks. The command is used as follows:

feelpp-benchmarking-exec --machine-config <machine-config> \
                            --benchmark-config <benchmark-config> \
                            --plots-config <plots-config> \
                            --custom-rfm-config <custom-rfm-config>

Additionally, the -v flag can be used to increase the verbosity, the --dry-run flag can be used to simulate the execution and only generate scheduler scripts without submitting them, and the --website flag can be used to render the final reports and launch an http server to visualize the generated dashboard.

The feelpp-benchmarking-exec command will export the ReFrame performance reports, along with a snapshot of the plots configuration used for the current benchmark, and a website_config.json file that contains the information necessary to render the dashboard.

There is also a feelpp-benchmarking-render command that is used to render the dashboard using Asciidoc and Antora, from a given website configuration, into a specified modules directory. Using this command allows to tweak figures and the dashboard layout, without having to re-run the benchmarks.

There are a ton of powerful features of feelpp.benchmarking that were not mentioned in this course, they will be mentioned on the advanced training session, such as working with containers, pruning the parameter space and handling memory (or other) constraints.