摘要： 针对传输触发架构下代码生成中指令调度的流水线冲突、调度死锁、资源冲突等问题，给出一种基于最小延时的遗传搜索算法模型，将软件旁路优化和资源动态分配优化整合到该模型中。实验结果表明，该算法能产生较高质量的并行代码，90%以上测试用例的指令级并行度高于表调度算法获得的结果。

关键词: 指令调度, 传算法, 输触发架构, 小延时, 发式搜索

Abstract: This paper proposes a genetic search algorithm model based on minimum delay to solve the instruction scheduling problems such as pipeline scheduling conflicts, deadlock scheduling and resource conflicts. It gives a genetic search algorithm model based on minimum delay. The software bypass optimization and dynamic allocation of resources are integrated into this model. Experimental results show that the algorithm can produce higher quality parallel code. Compared with the list scheduling algorithm, instruction level parallelism of more than 90% testbench is higher by using this algorithm.

Key words: instruction schedule, Genetic Algorithm(GA), Transport Triggered Architecture(TTA), minimum delay, euristic search

中图分类号: 

• N945.15