Compiler-Support for Robust Multi-core Computing

Abstract

Embedded computing is characterised by the limited availability of computing resources. Further, embedded systems are often used in safety-critical applications with real-time constraints. Thus, the software development has to follow rigorous procedures to minimise the risk of system failures. However, besides the inherent application complexities, there is also an increased technologybased complexity due to the shift to concurrent programming of multi-core systems. For such systems it is quite challenging to develop safe and resourceefficient systems. In this paper we give a plea for the need of better software development tools to cope with this challenge. For example, we outline how compilers can help to simplify the writing of fault-tolerant and robust software, which keeps the application code more compact, comprehensive, and maintainable. We take a rather extreme stand by promoting a functional programming approach. This functional programming paradigm reduces the complexity of program analysis and thus allows for more efficient and powerful techniques. We will implement an almost transparent support for robustness within the SAC research compiler, which accepts a C-like functional program as input. Compared to conventional approaches in the field of automatic software-controlled resilience, our functional setting will allow for lower overhead, making the approach interesting for embedded computing as well as for high-performance computing. The research leading to these results has received funding from the IST FP-7 research project "Asynchronous and Dynamic Virtualization through performance ANalysis to support Concurrency Engineering (ADVANCE)".