JCSE, vol. 8, no. 1, pp.34-42, 2014

DOI: http://dx.doi.org/10.5626/JCSE.2014.8.1.34

Exploiting Standard Deviation of CPI to Evaluate Architectural Time-Predictability

Wei Zhang and Yiqiang Ding
Department of Electrical and Computer Engineering, Virginia Commonwealth University, Richmond, VA, USA

Abstract: Time-predictability of computing is critical for hard real-time and safety-critical systems. However, currently there is no metric available to quantitatively evaluate time-predictability, a feature crucial to the design of time-predictable processors. This paper first proposes the concept of architectural time-predictability, which separates the time variation due to hardware architectural/microarchitectural design from that due to software. We then propose the standard deviation of clock cycles per instruction (CPI), a new metric, to measure architectural time-predictability. Our experiments confirm that the standard deviation of CPI is an effective metric to evaluate and compare architectural time-predictability for different processors.

Keyword: Performance; Reliability; Hard real-time systems; WCET analysis; Time-predictability

Full Paper:   168 Downloads, 2317 View