Permanent Data Structure Processing

Each run of real data is proceeded by a special record: the Run Header. This data structure holds information that is global to the entire run that follows. If a file contains multiple runs, there will be multiple headers. Some, but not all, Monte Carlo runs can also start with a generated Run Header. Other records are also permanent, in the sense that they apply to events beyond the current one although, in general, such records will not apply to an entire run. So a single run would normally contain several such records. Run Headers and other permanent records always preceed the data to which they apply. Permanent records of all types are processed in the following way:-

• When generated or input, they are merged into the permanent data structure in division 2 of the Q store to ensure that they persist throughout the run. Any parts of the data structure that become obsolete are removed.

  For an explanation about divisions in ZEBRA memories see:-

  Companion/ ZEBRA: A Beginner's Guide / Stores and Divisions

  Each time a new run header is encountered, it replaces any previous one already in division 2, but for some other permanent data a previous is kept to allow for interploation as appropriate.

• Most processors simply ignore permanent data, however the input processor reads them, and the output processor writes them.

This scheme ensures that all permanent data is transmitted through SNOMAN; they appear on the output file if they were present on the input file. The copy kept in division 2 is available to any code that requires it at any point throughout the run.