In google benchmark: there is a Iterms_per_seconds result and we can use the fixture way to test the bench.
items per second is the throughput. items is defined by the benchmark author and is completely optional. the benchmark author can also define bytes processed for bytes per second, if that is more meaningful.
you don't need fixtures, but they provide a way to do one-off setup and teardown.