Release date: March 2015. Confidential.
This document reports on the activities conducted in Task 4.2. It summarizes the approach taken in the project to estimate the worst-case execution time (WCET) of an OpenMP 4 task, based on an individual analysis of the Worst-Case Execution Time (WCET) of all its parts.
First, we give an overview of the Kalray MPPA-256 platform and describe the existing models to describe the architecture. These models are typically use in static timing analysis methods to derive a WCET estimate of a sequential code. Next, we summarize the state-of-the-art WCET timing analysis techniques, including static, measurement-based, hybrid, and probabilistic approaches. We discuss their up and downsides and justify why those already-available techniques can hardly be used in the context of P-SOCRATES.
Then, we introduce our new measurement-based approach. We explain how we collect the execution traces at runtime and how we process those traces to derive a WCET estimate for each part of the OpenMP task under analysis. Finally, we explain our higher-level analysis to produce a WCET estimate of the task based on the WCET estimates of all its parts.