The Predictable Execution Model in Practice

Adoption of multi- and many-core processors in real-time systems has so far been slowed down, if not totally barred, due do the difficulty providing analytical guarantees on worst-case execution times. The Predictable Execution Model (PREM) proposed to solve this problem, but its practical support requires significant code refactoring, a task better suited for compilation tool chain than human programmers. Implementing PREM compiler presents challenges conform requirements, such as guaranteed upper bounds memory footprint generation efficient schedulable non-preemptive regions. This article comprehensive description how can be implemented, based several years experience from community. We provide accumulated insights best balance conformance requirements performance present novel techniques that extend applicability simple benchmark suites real-world applications. show transformed by enables timing predictable modern commercial off-the-shelf hardware, protect 99.4% accesses random replacement policy caches at only 16% loss benchmarks PolyBench suite. Finally, we imposed programming model are well-aligned with current coding guidelines critical software, promoting easy adoption.