Optimal Design Parameter Exploration of Real-Time Software and its Application to Embedded Software Reuse

在本研究中，我们开发了一种多种调度策略共享处理/通信资源的多任务实时嵌入式软件的性能验证方法，以及一种满足性能要求的设计参数优化方法。在开发的验证方法中，我们使用优先秒表Petri网（PrSwPN）验证给定的性能要求是否得到满足。在所开发的参数优化中，在满足性能要求的情况下自动放宽每个时间预算（每个任务分配的执行时间），然后对每个任务的（可能递归的）程序进行静态分析，得出循环迭代次数和处理器频率等设计参数的条件，并使用数学规划对其进行自动优化。这些方法使得用低成本/低功耗资源开发要求相同性能的嵌入式系统，以及在不同的执行平台中重用软件组件变得更加容易。

In this research, we developed a performance verification method for multitask real-time embedded software with processing/communicating resources shared by various scheduling policies, and a design parameter optimization method satisfying performance requirements. In the developed verification method, we verify whether a given performance requirements are satisfied using Prioritized Stopwatch Petri Nets(PrSwPN). In the developed parameter optimization, each time budget(an allocated execution time for each task) are automatically relaxed while satisfying performance requirements, and then the(possibly recursive) program for each task is statically analyzed to derive a condition of design parameters such as loop iteration numbers and a processor frequency for automatically optimizing them using mathematical programming. These methods make it easier to develop embedded systems which requires the same performance with low cost/low power resources, and to reuse of software components into different execution platforms.