Research on reflective computation model based on data-parallelism

本文提出了一种用数据并行语言C^* 在连接机CM-5上实现并行逻辑语言GHC子集的方法。我们的目标是提高大规模并行计算机的可编程性和MIMD语言的可移植性。在我们的实现中，我们关注两种并行性，即，目标和数据结构的并行性。目标和数据都被分配给C^* 的虚拟处理器。每个目标被分配给一个虚拟处理器，目标以同步并行的方式执行。另外，每个数据都被分配给一个虚拟处理器，因此，可以同时访问。在反射，我们提出了反射GHC在BinProlog中的新实现。在反射式GHC的实现中，Blackbooad的概念和对BinProlog二进制子句的转换被证明是非常有效的，利用BinProlog的这些特性，我们提出了反射式GHC的新实现。

We proposed an implemetation method for a subset of parallel logic language GHC,using data parallel language C^* on connection machine CM-5. Our objectives were to improve the programmability of massive parallel computers and the portability of MIMD languages.In our implementation, we focused two kinds of parallelism, i.e., the parallelism of goals and that of data structures. Both of goals and data are assigned to virtual processors of C^*. Each goal is assigned to a virtual processor and the goal is executed synchronized parallel manner. Also each data is assigned to a virtual processor, and thus, can be accessed simultaneously.In relate to Reflection, we proposed the new implementation of Reflective GHC in BinProlog. The concept of blackbooad and the translation to the binary clause of BinProlog turned out to be very effective in implementing Reflective GHC.By using those features of BinProlog, we proposed the new implementation of Reflective GHC.