A Fine-Grained Analysis for Parallel Tasks with Spin-Locks under Global Fixed-Priority Scheduling

The synchronization of parallel tasks is critical for the schedulability of multiprocessor real-time systems. While non-preemptive spin-locks are widely adopted for their simplicity, analyzing the worst-case blocking time under the more dynamic global scheduling remains challenging. The primary difficulty stems from transitive delays induced by non-preemptive busy-waiting, and existing techniques relying on the traditional coarse-grained resource model can be pessimistic. This paper presents an improved blocking analysis framework for parallel tasks under global fixed-priority (GFP) scheduling. Our approach incorporates a fine-grained resource model that explicitly accounts for both per-request lengths and request allocations. By leveraging a bi-directional bounding technique within a linear programming formulation, the proposed analysis framework ensures a tighter bound on transitive delay and achieves the property of inflation-freeness. Evaluations on randomly generated task sets demonstrate that the proposed analysis outperformed the state-of-the-art in terms of schedulability.