Worst-Case Performance for Real-Time Systems

Author(s): Martin Schoeberl, Erling Jellum, Shaokai Lin, Chadlia Jerad, Emad Jacob Maroun, Marten Lohstroh, and Edward A Lee

Abstract
Real-time systems need to be built out of tasks for which the worst-case execution time is known. To enable accurate estimates of worst-case execution time, some researchers propose to build processors that simplify that analysis. These architectures are called precision-timed machines or time-predictable architectures. However, what does this term mean? This paper explores the meaning of time predictability and how it can be quantified. We show that time predictability is hard to quantify. Rather, the worst-case performance as the combination of a processor, a compiler, and a worst-case execution time analysis tool is an important property in the context of real-time systems. Note that the actual software has implications as well on the worst-case performance. We propose to define a standard set of benchmark programs that can be used to evaluate a time-predictable processor, a compiler, and a worst-case execution time analysis tool. We define worst-case performance as the geometric mean of worst-case execution time bounds on a standard set of benchmark programs.

Electronic Downloads

• https://doi.org/10.1017/cbp.2025.1

---

Citation Formats

• APA
  

                  
  Martin Schoeberl, Erling Jellum, Shaokai Lin, Chadlia Jerad, Emad Jacob Maroun, Marten Lohstroh, and Edward A Lee. (2025). Worst-Case Performance for Real-Time Systems (Master's thesis). In Research Directions: Cyber-Physical Systems                      
                  
                  

• MLA
  

                  
  Martin Schoeberl, Erling Jellum, Shaokai Lin, Chadlia Jerad, Emad Jacob Maroun, Marten Lohstroh, and Edward A Lee. "Worst-Case Performance for Real-Time Systems." Master's thesis, 2025. Research Directions: Cyber-Physical Systems,                       
                  
                  

• Chicago
  

                  
  Martin Schoeberl, Erling Jellum, Shaokai Lin, Chadlia Jerad, Emad Jacob Maroun, Marten Lohstroh, and Edward A Lee. "Worst-Case Performance for Real-Time Systems." Master's thesis, 2025. In Research Directions: Cyber-Physical Systems                      
                  
                  

• BibTeX
  

                      
  @article{SchoeberlEtAl:25:RealTime,
  	author = {Martin Schoeberl, Erling Jellum, Shaokai Lin, Chadlia Jerad, Emad Jacob Maroun, Marten Lohstroh, and Edward A Lee},
  	title = {Worst-Case Performance for Real-Time Systems},
  	journal = {Research Directions: Cyber-Physical Systems},
  	type = {Master's thesis},
  	month = {January},
  	year = {2025},
  	abstract = {Real-time systems need to be built out of tasks for which the worst-case execution time is known. To enable accurate estimates of worst-case execution time, some researchers propose to build processors that simplify that analysis. These architectures are called precision-timed machines or time-predictable architectures. However, what does this term mean? This paper explores the meaning of time predictability and how it can be quantified. We show that time predictability is hard to quantify. Rather, the worst-case performance as the combination of a processor, a compiler, and a worst-case execution time analysis tool is an important property in the context of real-time systems. Note that the actual software has implications as well on the worst-case performance. We propose to define a standard set of benchmark programs that can be used to evaluate a time-predictable processor, a compiler, and a worst-case execution time analysis tool. We define worst-case performance as the geometric mean of worst-case execution time bounds on a standard set of benchmark programs.},
  	URL = {https://doi.org/10.1017/cbp.2025.1}}