Coherence-Aware Task Graph Modeling for Realistic Application

As multicore systems continue to scale, cache coherence has emerged as a critical determinant of system performance, with coherence behavior and task execution deeply intertwined—reshaping intertask dependencies. Task graph modeling offers a structured way to capture such dependencies and serves as the foundation for many system-level design strategies. However, these strategies typically rely on predefined task graphs, while many real-world applications lack explicit task graphs and exhibit dynamic, data-dependent behavior, limiting the effectiveness of static approaches. Thus, many task graph modeling methods for realistic workloads have been developed. However, they either rely on implicit methods—using application-specific features without producing explicit graphs—or generate graphs tailored to fixed scheduling models, offering limited generality. Critically, they overlook coherence interactions, resulting in a mismatch between design assumptions and actual runtime behavior. To address these limitations, we propose CoTAM, a Coherence-Aware Task Graph Modeling framework for realistic workloads that constructs a unified task graph reflecting runtime behavior. CoTAM analyzes the impact of coherence by decoupling its behavior from overall execution. It then quantifies the influence of coherence through a learned weighting scheme, infers inter-task dependencies for coherence-aware task graph generation. Extensive experiments demonstrate the superiority of CoTAM over implicit methods, not only bridging the gap between dynamic workload behavior and existing designs, but also underscoring the importance of incorporating cache coherence into task graph modeling for accurate and generalizable system-level analysis.