摘要
虽然OpenMP是多核体系结构上的流行多线程并行编程模型,但是OpenMP编译器不检查数据相关性、访问冲突和其他可能导致程序错误执行的问题,这些问题传统上完全依赖用户使用锁机制来保证程序的正确性.锁机制的并行编程中存在并行程序效率和并行编程难度的矛盾.粒度大的锁机制编程容易,可应用的并行性挖掘比较差;粒度小的锁机制应用的并行性挖掘较好,可编程难度大,容易带来优先权倒置、死锁和锁护航等问题.通过动态二进制插桩技术,扩充OpenMP支持事务存储执行功能,可有效缓解OpenMP并行编程中并行程序效率和并行编程难度之间矛盾.
Although OpenMP is the popular multithread programming model on CMP architecture, OpenMP compilers do not check data dependency, memory access confliction and other problems likely to cause program errors. The traditional lock is applied by programmers to guarantee the correctness of their programs. It is easy to write coarse-grain lock programs, but the parallelism of the program may be lost. On the other hand, potential parallelism of a programs can be found by writing fine-grain lock programs, but it may bring about unwanted problems, such as priority inversion, deadlock, etc. Applying binary instrumentation technology to realize the extension of OpenMP to support transactional memory can effectively alleviale the contradiction between the simplicity and productivity in writing parallel OpenMP programs.
基金
Supported by the National Natural Science Fund Key Project(60533020)
Natural Science Found of Anhui Province(090412068)
Major Project of National Science and Technology(2009ZX01034-001-001-002)
