Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to upgrade your browser.

Log In
Sign Up

Figure from:An Efficient Real-Time Multiprocessor Scheduling Algorithm

See full PDF downloadDownload figure

Figure 2. (c) The handleEorZEvent procedure After handling all E events, the procedure then continues to handle all Z events (lines 17-27). In (line 18) the procedure checks heap Hy for any tasks whose laxity became zero. If such task exists, then it will preempt the task with the minimum remaining execution time from heap He after updating its key to the next time when it will fire an E event (lines 20, 22-26). The preempted task from heap He will be added to heap Hy after setting its key to the next time when it will fire a Z event (lines 19, 21, 27). — Figure 2 (c) The handleEorZEvent procedure After handling all E events, the procedure then continues to handle all Z events (lines 17-27). In (line 18) the procedure checks heap Hy for any tasks whose laxity became zero. If such task exists, then it will preempt the task with the minimum remaining execution time from heap He after updating its key to the next time when it will fire an E event (lines 20, 22-26). The preempted task from heap He will be added to heap Hy after setting its key to the next time when it will fire a Z event (lines 19, 21, 27).

Related Figures (10)

In our new approach, we propose three states for each task, as depicted in Figure | below, Running, Waiting and Not Active. The tasks traverse between the states with th advance of time.

The main procedure, as shown in Figure 2 (a) below, starts with initializing the system by calling the initialize procedure (line 1). Unlike LRE-TL, which calls the initialize procedure at the end of each TL-plane, in our algorithm this procedure is called once when the system is started to initialize the heaps. After the initialize procedure ends, the system is started by checking for all types of events as the time advanced (lines 3-6). Firstly the system checks for both E or Z events (line 3) and whenever one of them occurs, the system respond by calling the handleEorZEvent procedure (line 4). After that, the system checks for the A event (line 5), and if a task became an active i.e. it is released, the system respond by calling the handleAEvent procedure (line 6). After that, all processors are instructed to perform their assigned jobs (line 7), and the system waits until the next time an event will be fired (line 8).

Figure 2. (b) The initialize procedure In the initialize procedure, which is shown in Figure 2 (b) below, a priority queue (pq) before they are assigned to the processors (lines 2-4) assumed to be sorted according to task’s laxity. A task with least laxity wil Then we extract the first m tasks from the priority queue and add them to their keys to the next time when they fire an E event i.e. they finish 1 tasks are added to a . The priority queue is have a higher priority. heap He after updating lines 6-12). All other remaining tasks will be added to heap Hy with their keys set to the next time they become critical (lines 14-15). This will ensures that the tasks selected for execution first y are the most critical ones. Such tasks will continue to execute until either they finish or another task(s) from heap Hy became critical (reaches zero laxity), as we will see later in the hadleEorZEvent Procedure. In lines (17-18) remaining processors, if any, are set to null.

4.8 Example critical, then it will be added to heap Hw after updating its key to the next time when it will fir a Z event (lines 21-23). If it is critical, then it will preempt the task with the minimum remainin execution from heap Hp (lines 24-27, 29-32). The preempted task will be added to heap Hy afte updating its key to the next time when it will fire a Z event (lines 28, 33).

Figure 4. Comparison of migrations per job between USG algorithm and the best known

Figure 5. Comparison of preemptions per job between USG algorithm and the best known algorithms

Figure 6. Average deadline misses per task for USG and G-EDF

Figure 7. Percentage of scheduled task sets for USG, LLF and G-EDF

Figure 8. Percentage of unscheduled task sets for USG, LLF and G-EDF

About
Press
Papers
Topics
We're Hiring!
Help Center

Find new research papers in:
Physics
Chemistry
Biology
Health Sciences
Ecology
Earth Sciences
Cognitive Science
Mathematics
Computer Science

Terms
Privacy
Copyright
Academia ©2025