Abstract not found

This paper presents the architecture of a virtual machine designed specifically for the execution of multiple languages, which we call Seam. The architecture consists of a number of generic components, usable by all languages, and of a number of interfaces for which implementations have to be provided by language implementors. Our contribution is the identification of the generic services and the clean design for the parameterization over their language-specific aspects. The goal of Seam is to provide both for ample reuse and simple language implementation, concerning both compilers and runtime components, and to be a platform for language interoperation. We have implemented a prototype version of Seam and validated it with two language implementations. We present a full running implementation of Alice and a nave implementation of a Java Virtual Machine running on Seam. The paper presents first implementation e#ort and performance results for the prototype.

Traditionally an execution environment faces a trade-off between providing high-level or low-level concurrency mechanisms. The former trades flexibility for ease-of-use, while the latter results in a concurrency management closer to applications needs at the cost of an increase in the complexity of the applications code. Thus, one way or another, an application programmer has to match his application's semantic to the set of abstractions exported by the target execution environment.