In this paper we describe a collection of techniques for the design and implementation of concurrent compilers. We begin by describing a technique for dividing a source program into many streams so that each stream can be compiled concurrently. We discuss several compiler design issues unique to concurrent compilers including source program partitioning, symbol table management, compiler task scheduling and information flow constraints. The application of our techniques is illustrated by a complete design for a concurrent Modula-2+ compiler. After describing the structure of this compiler's performance that demonstrates that significant improvements in compilation time can be achieved through the use of concurrency.

Traditional compilers are usually sequential programs that serially process source programs through lexical analysis, syntax analysis, semantic analysis and code generation. The availability of multiprocessor computers has made it feasible to consider alternatives to this serial compilation process. The authors are currently engaged in a project to devise ways of structuring compilers so that they can take advantage of modern multiprocessor hardware. This paper is about the most difficult aspect of concurrent compilation: concurrent semantic analysis.