Marianne De Michiel

MBBA: A Multi-Bandwidth Bus Arbiter for Hard Real-Time

Multi-core architectures are being increasingly used in embedded systems as they offer several advantages: improved hardware integration, low thermal dissipation and reduced energy consumption, while they make it possible to improve the computing power. In order to run real-time software on a multicore architecture, computing the Worst-Case Execution Time of every thread should be achievable. This notably involves bounding memory latencies by employing a predictable bus arbiter. However, state-of-the-art techniques prove to be irrelevant to schedule unbalanced workloads in which some threads require more bus bandwidth than the other ones. This paper proposes a new bus arbitration scheme that ensures that the shared bus latencies can be upper bounded. Compared to other schemes that make the bus latencies predictable, like the Round-Robin protocol, our approach defines several levels of bandwidth to meet requirements that may vary from one thread to another. Experimental results (WCET estimates) show that the worst-case bus latency is noticeably shortened, compared to Round-Robin, for the cores with highest priority that get the largest bandwidth. The relevance of the scheme is shown through an example workload composed of various benchmarks.

FFX: a portable WCET annotation language

In order to ensure safety of critical real-time systems it is crucial to verify their temporal properties. Such a property is the Worst-Case Execution Time (WCET), which is obtained by architecture-dependent timing analysis and architecture-independent flow fact analysis. In this article we present a WCET annotation language which is able to express such information originating from the user or the analysis. The open format, named FFX to stand for Flow Facts in XML, is portable, expandable and easy to write, understand and process. We argue that FFX allows to reuse and exchange the annotation files among WCET tools. FFX therefore permits to tighten WCET results and decreases the effort to support new architectures. Additionally, FFX flow fact files allow fair comparisons of both flow facts and WCET results. FFX can be used for quality assurance when developing new analysis techniques, using it as a flow fact database to test against. We present a small case study exemplifying the above points. Our case study puts special focus on the aspect of comparability and information exchange among WCET tools. In our experiments with FFX, we use the WCET analysis tool chains Otawa/oRange and r-TuBound/CalcWCET167.

Partial flow analysis with oRange

In order to ensure that timing constrains are met for a Real-Time Systems, a bound of the Worst-Case Execution Time (WCET) of each part of the system must be known. Current WCET computation methods are applied on whole programs which means that all the source code should be available. However, more and more, embedded software uses COTS (Components ...), often afforded only as a binary code. Partialisation is a way to solve this problem. In general, static WCET analysis uses upper bound on the number of loop iterations. oRange is our method and its associated tool which provide mainly loop bound values or equations and other flow facts information. In this article, we present how we can do partial flow analysis with oRange in order to obtain component partial results. These partial results can be used, in order to compute the flow analysis in the context of a full application. Additionally, we show that the partial analysis enables us to reduce the analysis time while introducing very little pessimism.

Expressing and Exploiting Conflicts over Paths in WCET Analysis.

The presence of infeasible paths in a program is a source of imprecision in the Worst-Case Execution Time (WCET) analysis. Detecting, expressing and exploiting such paths can improve the WCET estimation or, at least, improve the confidence we have in estimation precision. In this article, we propose an extension of the FFX format to express conflicts over paths and we detail two ways of enhancing the WCET analyses with that information. We demonstrate and compare these techniques on the Malardalen benchmark suite and on C code generated from Esterel.

Normalisation of Loops with Covariant Variables

Temporal property verification is utterly important to ensure safety of critical real-time systems. A main component of this verification is the computation of Worst Case Execution Time (WCET) that requires, in turn, the determination of loop bounds. Although a lot of efforts have been performed in this domain, it remains relatively common cases which are unsolved. For example, to our knowledge, no fast automatic method can cope with the loop bound of a simple binary search look-up. In this paper, we present an approach to solve such loops by using arithmetico-geometric series, that is, loops with arithmetic and/or geometric incrementation with several variables. We have implemented and experimented this approach in our tool oRange.

Working Around Loops for Infeasible Path Detection in Binary Programs

The research of a safe Worst-Case Execution Time (WCET) estimation is necessary to build reliable hard, critical real-time systems. Infeasible paths are a major cause of overestimation of theWorst-Case Execution Time (WCET): without data flow constraints, static analysis by implicit path enumeration will take into account semantically impossible, potentially expensive execution paths, making theWorst-Case Execution Path unreachable in practice. We present in this paper an approach that allows to significantly tighten the WCET by identifying infeasible paths, namely in loops, and injecting them as additional Integer Linear Programming (ILP) constraints during the WCET computation. Our entire analysis, albeit platform independent, works directly on binary programs in order to get the tightest, most reliable WCET. Impactful infeasible paths are largely found within (often nested) loops; therefore having an efficient, exploitable and reasonably scalable representation of the state of a program within loops is a key challenge of infeasible path analysis. We show ours to yield decidedly significant results on a selection of benchmarks from actual hard real-time applications as well as the classic M¨alardalen suite.