模型驱动的开发异构多核多级并行的方法

将异构多核处理器的资源分配问题抽象为一个三维优化空间，空间中的每个元素对应多任务程序的一个并行策略，确定程序有效并行策略的问题转化为优化空间的搜索问题。为了加速搜索，提出一种基于遗传算法的智能搜索方法。实验表明，模型驱动的并行策略选择方法能够在极短的时间内为程序生成较好的并行策略。

Model-driven approach to exploiting multi-level parallelization on heterogeneous multi-cores

The resources allocation for heterogeneous multi-cores is modeled as a three dimensional optimization space. Each element in the optimization space corresponds to a parallelization scheme. Finding efficient parallelization schemes for a multi-task program is to explore this search space. In order to accelerate space exploration, an intelligent search based on genetic algorithm is adapted. Experiments show that the model-driven approach can derive good parallelization schemes in very short time.