Jackal, A Compiler Based Implementation of Java for Clusters of Workstations

This paper describes the design of Jackal, a compiler-driven distributed shared memory implementation of the Java programming language. Our goal is to efficiently execute (unmodified) multithreaded Java programs on a cluster of workstations. Jackal consists of a native Java compiler and a runtime system that implements a distributed shared memory protocol for variable sized memory regions. The Jackal compiler stores Java objects in shared regions and augments the programs it compiles with access checks. These access checks drive the memory consistency protocol. The Jackal compiler implements several optimizations to reduce the overhead of these software access checks. The main contributions of this paper are: techniques to allow a (Java) compiler to target a fine-grained allsoftware DSM using access-checks and compiler optimizations to achieve good efficiency.