A memory management method in task for heterogeneous multicore processors to realize efficient execution

本文提出了一种新的SIMD并行化任务中的内存管理方法，以获得异构多核处理器的实际能力，并使执行时间长的应用程序快速。我们实现了该方法，并通过使用一个实际的应用程序，LSI测试代码进行评估。我们已经完成了所提出的方法的基本实现和评估。我们表明，所提出的方法执行应用的LSI测试代码具有大的测试数据与原始代码的速度的两倍。

This study proposes a new memory management method in tasks for SIMD parallelization in order to derive the practical capability of heterogeneous multicore processors and to make applications that take long time execution time fast. We implement the method and evaluate it by using a practical application, a LSI test code. We have finished a basic implementation and evaluations of the proposed method. We show the proposed method executes the applied LSI test code having large test data with twice the speed of the original code.