Abstract: Intelligent production-logistics operation control and decision-making in complex and dynamic environments is one of the core technologies in discrete manufacturing systems. To resolve the complexity originate from system structures and multiple resource interdependencies, complexity of production-logistics operations under various disturbances and multiple resource constraints, and the difficulty in bridging the gap between decision-making and actual execution under uncertainty, this paper, inspired by the “Out-of-Order execution” in high-performance CPU, proposes an system architecture for instruction-based production-logistics operations towards smart manufacturing. This architecture consists of dynamic deconstruction of manufacturing systems, spatiotemporal reasoning of uncertainty, redefined operational and interactive logic, and core decision and control methods, reshaping the production-logistics operation mode for smart manufacturing systems. Case study is carried out to verify the performance of out-of-order execution in stochastic environments with frequent disturbances. The results show that out-of-order execution performs well in terms of makespan, average job tardiness, average job flow time, and average operator utilization rate, exhibiting smaller fluctuations and more stable performance in stochastic environments. Lastly, the future perspectives of “out-of-order execution” in smart manufacturing systems are discussed, providing new ideas and methods for optimizing production-logistics operations in smart manufacturing factories.