Abstract

any impure-LISP program running in time t can be compiled into a pure-LISP program running in time O(t log t): first implement the impure-LISP operations using an array of size at most t. Then represent the array as a balanced binary tree, which can be done in pure LISP. The main result of the paper is a lower-bound theorem. It can roughly be described as follows. A problem P is presented, that can be solved in linear time, t = O(n), in impure LISP. It is proved that for any pure-LISP program p for P , the worst-case time complexity is \Omega\Gamma n log n). 2 Restrictions of the Proof and Open Problems The lower-bound result requires two restrictive assumptions. We first describe the restrictions and their technical implications. Next, we discuss the two questions that

Citations