Propagation is at the very core of t can provide signi: it can provide significant performance boosts as long as the search space reduction is not outweighed by the cost for running the propagators. A lot of research effort in the CP community is directed toward improving this trade-off. While experimental evaluation is here of the greatest importance, there exists no systematic and flexible methodology to measure the exact benefits provided by a given (new) filtering procedure. This work proposes such a framework by relying on replaying search trees to obtain more realistic assessments. Reducing propagation overhead is done chiefly by 1) devising more efficient algorithms or by 2) using on-line control policies to limit the propagator activations, i.e., mechanisms to reduce the number of propagator calls. In both cases, obtaining improvements is a long and demanding process with uncertain outcome. We propose a method to assess the potential gain of both approaches before actually starting the endeavor, providing the community with a tool to best direct the research efforts. In order to visualize benefits of actual global constraints and the potential of their improvement, we suggest the use of performance profiles. Our approach is showcased for well-known global constraints: alldifferent, cumulative, binpacking and unary (with transition times).

中文翻译：

---

在实践中，全球约束的效率如何？

传播是t可以提供的重要核心：只要不减少传播空间的运行成本，就可以大大提高性能。CP社区中的许多研究工作都旨在改善这种平衡。尽管此处的实验评估最为重要，但尚无系统，灵活的方法来测量给定（新）过滤程序所提供的确切收益。这项工作依靠重播提出了这样一个框架搜索树以获得更现实的评估。减少传播开销的主要方法是：1）设计更有效的算法，或2）使用在线控制策略限制传播者的激活，即减少传播者调用次数的机制。在这两种情况下，获得改进都是一个漫长而艰巨的过程，并且结果不确定。我们提出了一种在实际开始尝试之前评估这两种方法的潜在收益的方法，从而为社区提供了最佳指导研究工作的工具。为了可视化实际全局约束的好处及其改进的潜力，我们建议使用性能概况。我们针对已知的全局约束展示了我们的方法：所有不同，累积，装箱和一元（具有转换时间）。