The buried binding and dead binding problems of Lisp 1.5: sources of incomparability in garbage collector measurements

Abstract

Lisp has become the language of choice for many applications such as artificial intelligence programs or symbol manipulation. The original implementation of Lisp 1.5 was a concise, elegant statement of the semantics of the language. Although production Lisp systems have undergone significant development and evolution since Lisp 1.5, including the development of sophisticated compilers, there have been few significant theoretical improvements in the implementations of these systems. Most improvements, such as arrays or shallow-binding, have been made more for the sake of speed than for the sake of storage. A notable exception to this is the technique of tail recursion, which can save more than just stack space.We believe that more can be done to reduce the storage requirements of Lisp programs. Although in many instances, the Lisp programmer can reduce the storage requirements of his program by deleting unneeded pointers as soon as possible, there is nothing he can do about systematic inefficiencies of the Lisp interpreter. This paper addresses itself to two sources of inefficiency in Lisp's variable binding mechanism--one of which is easy to detect---which prevent storage from being garbage collected long after its last reference. Implementations of Lisp which eliminate these situations should result in more economical execution for almost all Lisp programs which use a lot of storage, due to a lighter marking load on the garbage collector.