To benchmark performance, switch the two `defineSequence` lines in https://gitlab.cern.ch/snippets/639#L41 and change the `DIGI` on the last line to `MDF` and adjust threads/slots/events. The stdout gives you a total throughput of the sequence you run, and by comparing you see the impact of changes. For a more sophisticated timing number in optimal throughput conditions, we need to adjust some more stuff, but let me worry about this.