An Efficient Task Serializer for Hard Real-Time TMO Systems

Abstract

Timeliness-guaranteed computing considering the worst case can hardly be accomplished at run time because of the uncertainty caused by resource sharing, asynchronous events and conflicting timing constraints between tasks. An easier way to reach it is to do some off-line analysis from design time with predictable characteristics of tasks. The TMO model supporting time-triggered (SpM) and message-triggered (SvM) member-threads is a well-known real-time object-oriented task model that aims the design-time guaranteed computing. In this paper, a new efficient task serializer that allocates the executions of time-triggered periodic tasks of multiple TMOs without delay and preemption to avoid resource conflicts is presented. The serializer tries to find scheduling scenarios in that each SpM is executed at its exact trigger-time without delay based on the given WCET. This is done by recasting the initial offsets of SpMs to avoid overlaps of executions. And several methods to handle SvMs according to the types of SvMs using the serializer are also proposed.