Configuration and scheduling of the LHCb trigger application

For Run 3 of the Large Hadron Collider, the final stage of the LHCb experiment’s high-level trigger must process 100 GB/s of input data. This corresponds to an input rate of 1 MHz, and is an order of magnitude larger compared to Run 2. The trigger is responsible for selecting all physics signals that form part of the experiment’s broad research programme, and as such defines thousands of analysis-specific selections that together comprise tens of thousands of algorithm instances. The configuration of such a system needs to be extremely flexible to be able to handle the large number of different studies it must accommodate. However, it must also be robust and easy to understand, allowing analysts to implement and understand their own selections without the possibility of error. A Python-based system for configuring the data and control flow of the Gaudi-based trigger application is presented. It is designed to be user-friendly by using functions for modularity and removing indirection layers employed previously in Run 2. Robustness is achieved by reducing global state and instead building the data flow graph in a functional manner, whilst keeping configurability of the full call stack.