Efficient iteration over remage filesΒΆ

Note

This step is only relevant for applications involving large simulation files, or where the reboost configuration file option is used.

Simulation files are often very large. This poses a challenge in processing the data, since it may not always be feasible to read the full simulated file into memory.

In addition, not every simulated event will lead to energy deposits in the detectors, and each simulated event will have multiple rows in the output file (corresponding to the different steps). This introduces a challenge in iterating over the files, since a naive iteration could cause a simulated event to be split between two chunks.

Finally, it is often desirable to split the initial remage files (for example to split the file by run), to do this it is necessary to be able to read the data only corresponding to a range of simulated event ids.

The geant4 event lookup map (or glm) is a solution to these issues. The glm is a LGDO.VectorOfVectors() with three fields:

  • g4_evtid: The index of the simulated event,

  • n_rows: The number of rows of the output file corresponding to this event,

  • start_row: The first row of the file.

Note

  • This object should be computed once per output table (detector).

  • Geant4 events that do not lead to rows in the output file are still included.

The glm can be build from the remage outputs using the reboost.build_glm.build_glm() function. A command line interface has also been written:

reboost build-glm -h

We have created a reboost.iterator.GLMIterator class to allow for easy iteration over the stp files using the glm.