A parallel programming system for various tightly coupled multi-core architectures

在这项研究中，我们开发了并行编程系统，使没有经验的并行编程用户能够享受到当今多核和多核体系结构的好处。(1)我们开发了一个提供矩阵和可变长度列表的并行骨架库。(2)我们为GPGPU程序设计开发了一个骨架式并行框架。(3)我们开发了一个通用而强大的基于半环上矩阵乘法的自动并行化框架。(4)我们开发了一个系统，使用户能够从锁或软件事务内存中静态地选择合适的同步机制。

In this research, we have developed parallel programming systems that enable inexperienced users of parallel programming to enjoy the benefits of today's multi-core and many-core architectures. The obtained results can be summarized as follows.(1)We have developed a parallel skeleton library that offers matrices and variable-length lists.(2)We have developed a skeletal parallel framework for GPGPU programming.(3)We have developed a general and powerful framework for automatic parallelization based on matrix multiplication over a semiring.(4)We have developed a system that enables the user to statically select a suitable synchronization mechanism from a lock or a software transactional memory.