Individual task accounting for multiprocessor systems when executing

Abstract

In a multiprocessor system, a table is created for each multitask job in the main memory. The table contains a first memory field for storing a task count value 'm', second memory fields respectively identified by successive task count values for storing data respectively indicating the execution start times of 'm' simultaneously executed tasks, and third memory fields respectively identified by successive task count values for storing the run times of 'm' or more simultaneously executed tasks. In response to the roll-in of a given task, the task count value is incremented by 1, and real time data indicated by a system timer is stored into the second memory field identified by the task count value, allowing the selected task to be processed. At the end of the processing of any task, a difference is derived between real time data indicated by the system timer and an execution start time stored in the second memory field identified by the task count value. The difference is then added to a run time previously stored in the third memory field identified by the task count value, to produce a sum. This sum is stored back into the third memory field overwriting the previous run time. The task count value in the first memory field is decremented by 1 to allow the task to be rolled out. Differences are derived between run times stored in successive third memory fields.