ATLAS Offline Software
Filtering Class Reference

Consumes multiple decision inputs and filters those that are labelled with the required decision IDs. More...

Collaboration diagram for Filtering:

Detailed Description

Consumes multiple decision inputs and filters those that are labelled with the required decision IDs.

HLT Step sequencer A RoRSeqFilter is a generic filter over collections of Decision objects. It is the first thing to run in a Step, and all RoRSeqFilter within a Step must run before the Step proper can run. It manages the inter-Step logic and gates the execution of all algorithms needed by the Filter's set of chains within the following Step.

Typically each input collection will correspond to a collection of reconstructed objects which have been subject to a hypothesis algorithm. There is hence a one-to-one mapping between each input Decision object and a physics object in another collection (link key: "feature"). The decision is additionally decorated with the chain-ID of all Hypothesis Tools which passed the object in whole/as part of their selection.

Execution of the RoRSeqFilter is managed by Control Flow, it will be scheduled once all hypothesis algorithms in the previous step which feed it have finished or are known to not run in the event. It attempts to read in all implicit read handles.

All successfully read handles are filtered with respect to m_chainsProperty. Only Decision objects which contain an affirmative decision from at least one of the Chain-IDs within m_chainsProperty will be duplicated into the output collection. All non-empty output collections will be written to their write handles.

If at least one (non-empty) write handle is written, the filter will report TRUE to the Scheduler, unlocking the next Step of execution for the set of Chains which utilise this Filter in their path through the HLT Control Flow graph. Otherwise, it will report a filter decision of FALSE and terminate.

Note: Algorithms and hypos can be placed under the control flow of more than one RoRSeqFilter, in this case, only one is required to return a positive filter decision to unlock the control flow for this set of algorithms and hypo.


The documentation for this class was generated from the following file: