As time predictability is critical to hard real-time systems, it is not only necessary to accurately estimate the worst-case execution time (WCET) of the real-time tasks but also desirable to improve either the WCET of the tasks or time predictability of the system, because the real-time tasks with lower WCETs are easy to schedule and more likely to meat their deadlines. As a real-time system is an integration of software and hardware, the optimization can be achieved through two ways: software optimization and time-predictable architectural support. In terms of software optimization, we fi rst propose a loop-based instruction prefetching approach to further improve the WCET comparing with simple prefetching techniques such as Next-N-Line prefetching which can enhance both the average-case performance and the worst-case performance. Our prefetching approach can exploit the program controlow information to intelligently prefetch instructions that are most likely needed. Second, as inter-thread interferences in shared caches can signi cantly a ect the WCET of real-time tasks running on multicore processors, we study three multicore-aware code positioning methods to reduce the inter-core L2 cache interferences between co-running real-time threads. One strategy focuses on decreasing the longest WCET among the co-running threads, and two other methods aim at achieving fairness in terms of the amount or percentage of WCET reduction among co-running threads. In the aspect of time-predictable architectural support, we introduce the concept of architectural time predictability (ATP) to separate timing uncertainty concerns caused by hardware from software, which greatly facilitates the advancement of time-predictable processor design. We also propose a metric called Architectural Time-predictability Factor (ATF) to measure architectural time predictability quantitatively. Furthermore, while cache memories can generally improve average-case performance, they are harmful to time predictability and thus are not desirable for hard real-time and safety-critical systems. In contrast, Scratch-Pad Memories (SPMs) are time predictable, but they may lead to inferior performance. Guided by ATF, we propose and evaluate a variety of hybrid on-chip memory architectures to combine both caches and SPMs intelligently to achieve good time predictability and high performance. Detailed implementation and experimental results discussion are presented in this dissertation.