Wide-Area and Distributed Computation Paradigm Based on Concurrent Logic Programming

点击翻译按钮获取中文摘要

Concurrent logic languages provide natural mechanisms for describing concurrent computation by introducing dataflow synchronization on single-assignment variables. Previous work on concurrent logic programming focused on efficient parallel computation. Our research, in contrast, focused on distributed, wide-area computation and studied (1) declarative programming under non-uniform computational environments and (2) static analysis for safe distributed programs. Our results can be summarized as follows :1. Seamless distributed implementation - We designed and implemented network-transparent distributed logical variables (single-assignment channels) to transparently interconnect KL1 programs running on non-uniform computational nodes using sockets. We also designed naming service used for the naming and resolution of distributed logical variables. An exception handling mechanism was also designed and implemented.2. Resource management under distributed environments - We designed and implemented a linear type system for concurrent logic programs that statically analyzes data sharing by logical variables. Furthemore, we proposed a novel view of concurrent computation in terms of resource passing, and designed a capability type system that unifies and generalizes mode systems and linearity systems.3. Interpreter technology for code mobility - We designed an intermediate code, called treecode, to realize code mobility between non-uniform nodes, and described its interpreter using Flat GHC.