The SET FILTER command creates a new filter definition. When a
    subsequent PLOT or TABULATE command is entered, the restrictions
    specified in the filter definition are applied to all data points
    in the performance data file. A data point passes a filter if it
    satisfies any one filter restriction in the filter. Only those
    data points that pass all defined filters are tallied in the
    histogram or table.

    By using multiple restrictions in a single filter, you OR the
    restrictions; if a data point passes any one restriction,
    it passes the filter. By using multiple filters, you AND
    restrictions; a data point must pass all defined filters to be
    tallied in a histogram or table.

    Filter restrictions that do not apply to a given data point are
    ignored. A file name restriction, for example, is ignored for
    page fault data because no file name is collected for a page
    fault data point.

    All the qualifiers listed defined below are positional. That is,
    if you place a given qualifier after "SET FILTER" but before the
    first parameter, then the qualifier affects the entire command.
    If you place the qualifier after a filter specification, then the
    qualifier affects only that filter specification.

    If you specify /MAIN_IMAGE=prog-unit, /STACK_DEPTH=n and
    /CUMULATIVE=n, they are applied in the following order: /MAIN_
    IMAGE, /STACK_DEPTH, then /CUMULATIVE. Therefore, if you specify
    the /MAIN_IMAGE=prog-unit and /STACK_DEPTH=n qualifiers, the
    Analyzer will search for the first stack PC value within the
    specified program unit. However, it will not use the data
    point in this address, but in the address n call frames on the
    stack from the main image address, when checking address filter
    restrictions. If you include /CUMULATIVE on that command, the
    same action occurs, but the Analyzer also uses the PC values in
    all the call frames below that (or in the first n call frames
    down if /CUM=n).