Refine RELFC counter in z/TPFDF Data Collection

z/TPFDF data collection counters were originally designed to track the calls to specific z/TPFDF central database routines. Each of these routines had a specific task, such as obtaining a core block or releasing a record. These counts generally reflected the number of calls to z/TPF macros, such as GETCC and RELFC. However, this approach results in several inaccuracies. For example, when a DBDEL ALL,NOKEY or DBCLS RELFC is issued in the application while z/TPFDF data collection is running, the DFLNK START macro at the beginning of the central database routine that releases the records and forward chains increases the RELFC count by one, but multiple RELFC calls will potentially be made in the routine. Similarly, many other z/TPF macro calls within the central database routines are not counted, or the data collection counters are unconditionally updated when a routine is called even though the actual macro was conditionally skipped in the routine.

In addition, all data collection rates (per second) reported are approximately 5% higher than the actual values. This is because the data collection routines use the leftmost fullword of the TOD clock value for the start and stop time to calculate the run time. In reality, 1.048576 seconds has passed for every second counted. For example, if data collection runs for exactly 5 minutes, the internal routines will calculate the runtime as 286 seconds instead of 300 seconds. This inaccuracy results in the data collection "per second" rates being about 5% too high. Even though this discrepancy is consistent across all rates, the numbers reported should be accurate.