综合模块选择、资源共享与任务调度的SoC设计方案搜索算法

SoC Design Exploration Algorithm Intertwining Module Selection,Resource Sharing and Scheduling

在SoC系统的设计中,设计方案将直接决定最终系统的性能和成本.提出一种综合模块选择、资源共享与任务调度的设计方案搜索算法.该算法巧妙地将模块选择、共享方案的搜索与表调度融合在一个分支定界过程中,严格搜索表调度策略下的最优设计方案;将搜索与调度同时进行,具有较快的搜索速度.对多种不同规模的任务图进行测试,并与现有方法进行实验对比的结果表明,对于中小规模的系统设计,该算法能够在很短的时间内找到设计空间中的最优方案.

In the design of SoC systems, the design scheme the performance and cost of the ultimate system directly determines. A design exploration algorithm is proposed by intertwining the module selection, the resource sharing and the task scheduling. The algorithm integrates these three problems into a branch-and-bound procedure, and searches for the best solution under the list scheduling strategy. The searching is performed concurrently with the scheduling, which leads to a fast search speed. Several tests are taken for various task graphs of different scales. The performance of this algorithm is compared with those existing methods. The experimental results indicate that, for systems with small or medium scale, the algorithm can find the best solution in the design space in a short time.