This setup function should return the stuff you need in the tests. Anything returned will be available via the DATA variable inside the test cases. Running the setup function is not part of the benchmark and it's run separately for each test case. To learn more, check out this more advanced example .
Global dependencies are available in the setup function and every test case.
Note: No statistical analysis is used to validate the results. The tests are run in parallel for 3 seconds (with a 500ms warmup) and then operations per second are calculated.