Compiling Parallel Lisp for a Shared Memory Multiprocessor

There is a commonly held idea that although Lisp is a good language for prototyping and software development, it is often too inefcient to be useful in actual implementation. Advances in compiling Lisp have begun to change this notion. However, the situation is less clear in the case of parallel Lisp dialects. In this paper, we report on our experience in implementing a compiler for a parallel Lisp dialect we called BaLinda Lisp for a shared memory multiprocessor. For a class of problems, our compiler was able to achieve performances on Lisp programs that are near, if not equal to, their imperative (C) equivalents. However, on other problems, the situation is less than ideal. We investigate this phenomenum and outline the future challenges in realizing a production strength parallel Lisp compiler.