Advancing Multicore Timing Analysis: Modeling Contention and Deriving WCET Bounds in Cyclic Executive Systems

Identificador:  TDX:4712

Handle: https://hdl.handle.net/20.500.11797/TDX4712

Autores:  Palomo Teruel, Xavier

Resumen:
This thesis addresses the challenge of determining Worst-Case Execution Time (WCET) in real-time multicore systems, crucial for safety-critical applications. It proposes two novel methodologies for WCET analysis: an Integer Linear Programming (ILP) formulation that models contention scenarios for system-wide guarantees, and an iterative approach offering practical, less conservative task-level bounds. Both methods are adaptable to various multicore platforms. To validate these, microbenchmarks are developed at the assembly level to stress hardware components, ensuring precise event generation like cache misses. Performance events are derived using performance monitoring counters (PMCs) and units (PMUs). For platforms without PMUs, a mathematical approach and a software solution leveraging debug features estimate these events. These techniques enhance timing predictability and safety in multicore systems, bridging theoretical analysis with practical implementation. The contributions aim to improve realtime system reliability across constrained industrial platforms.