Multithreaded Reconstruction

The ordinary jet_reconstruct and tiled_jet_reconstruct interfaces return an independently owned ClusterSequence. They are the recommended interfaces for ordinary, single-threaded reconstruction: their allocation overhead is small, and the returned result can be retained freely.

High-throughput multi-threaded reconstruction

High-throughput multi-threaded applications that reconstruct many independent events in parallel can reuse N2Tiled storage with an N2TiledWorkspace. This avoids most per-event temporary allocations, which greatly improves performance. The cost of those allocations becomes more severe in multithreaded applications.

workspace = N2TiledWorkspace()

for event in events
    with_n2tiled_reconstruction(
        workspace,
        event;
        algorithm = JetAlgorithm.AntiKt,
        R = 0.4,
    ) do clusterseq
        # This selection is independently owned and can outlive the callback.
        jets = inclusive_jets(clusterseq; ptmin = 5.0)

        # To retain the complete clustering sequence instead:
        # retained_clusterseq = deepcopy(clusterseq)
    end
end

The callback receives a complete ClusterSequence, so history, constituent, and exclusive-jet queries remain available. Its jets and history borrow storage from the workspace and are overwritten by the next reconstruction using that workspace. Copy any data that must outlive the callback or use an ordinary owning interface instead.

Concurrent use

A workspace must not be shared concurrently or used reentrantly. Create one workspace for each long-lived worker task and keep that workspace owned by the task. The multithreaded N2Tiled example shows dynamic scheduling of events in parallel while preserving that ownership.