An Integrated Path and Timing Analysis Method based on Cycle-Level Symbolic Execution

  • Authors:
  • Thomas Lundqvist;Per Stenström

  • Affiliations:
  • Department of Computer Engineering, Chalmers University of Technology, SE–412 96 Göteborg, Sweden;Department of Computer Engineering, Chalmers University of Technology, SE–412 96 Göteborg, Sweden

  • Venue:
  • Real-Time Systems
  • Year:
  • 1999

Quantified Score

Hi-index 0.01

Visualization

Abstract

Previouslypublished methods for estimation of the worst-case executiontime on high-performance processors with complex pipelines andmulti-level memory hierarchies result in overestimations owingto insufficient path and/or timing analysis. This does not onlygive rise to poor utilization of processing resources but alsoreduces the schedulability in real-time systems. This paper presentsa method that integrates path and timing analysis to accuratelypredict the worst-case execution time for real-time programson high-performance processors. The unique feature of the methodis that it extends cycle-level architectural simulation techniquesto enable symbolic execution with unknown input data values;it uses alternative instruction semantics to handle unknown operands.We show that the method can exclude many infeasible (or non-executable)program paths and can calculate path information, such as boundson number of loop iterations, without the need for manual annotationsof programs. Moreover, the method is shown to accurately analyzetiming properties of complex features in high-performance processorsusing multiple-issue pipelines and instruction and data caches.The combined path and timing analysis capability is shown toderive exact estimates of the worst-case execution time forsix out of seven programs in our benchmark suite.