Techniques for reasoning about extensional properties of functional programs are well understood, but methods for analysing the underlying intensional or operational properties have been much neglected. This paper begins with the development of a simple but useful calculus for time analysis of non-strict functional programs with lazy lists. One limitation of this basic calculus is that the ordinary equational reasoning on functional programs is not valid. In order to buy back some of these equational properties we develop a non-standard operational equivalence relation called cost equivalence, by considering the number of computation steps as an `observable' component of the evaluation process. We define this relation by analogy with Park's definition of bisimulation in CCS. This formulation allows us to show that cost equivalence is a contextual congruence (and thus is substitutive with respect to the basic calculus) and provides useful proof techniques for establishing cost-equivalen...

The expense of developing and maintaining software is the major obstacle to the routine use of parallel computation. Architecture independent programming offers a way of avoiding the problem, but the requirements for a model of parallel computation that will permit it are demanding. The BirdMeertens formalism is an approach to developing and executing data-parallel programs; it encourages software development by equational transformation; it can be implemented efficiently across a wide range of architecture families; and it can be equipped with a realistic cost calculus, so that trade-offs in software design can be explored before implementation. It makes an ideal model of parallel computation. Keywords: General purpose parallel computing, models of parallel computation, architecture independent programming, categorical data type, program transformation, code generation. 1 Properties of Models of Parallel Computation Parallel computation is still the domain of researchers and those ...

A systematic approach is given for deriving incremental programs from non-incremental programs written in a standard functional programming language. We exploit a number of program analysis and transformation techniques and domain-specific knowledge, centered around effective utilization of caching, in order to provide a degree of incrementality not otherwise achievable by a generic incremental evaluator. 1 Introduction Incremental programs take advantage of repeated computations on inputs that differ only slightly from one another, avoiding unnecessary duplication of common computations. Given a program f and a certain input change \Phi, a program f 0 that computes the value of f(x \Phi y) efficiently by making use of the value of f(x) is called an incremental version of f under \Phi. The parameter y can be regarded as a change ffix to the input x. Methods of incremental computation have widespread applications, e.g., loop optimizations in optimizing compilers [1, 24, 9, 10] and ...

This paper describes a general approach for automatic and accurate time-bound analysis. The approach consists of transformations for building time-bound functions in the presence of partially known input structures, symbolic evaluation of the time-bound function based on input parameters, optimizations to make the overall analysis efficient as well as accurate, and measurements of primitive parameters, all at the source-language level. We have implemented this approach and performed a number of experiments for analyzing Scheme programs. The measured worst-case times are closely bounded by the calculated bounds. 1 Introduction Analysis of program running time is important for real-time systems, interactive environments, compiler optimizations, performance evaluation, and many other computer applications. It has been extensively studied in many fields of computer science: algorithms [20, 12, 13, 41], programming languages [38, 21, 30, 33], and systems [35, 28, 32, 31]. It is particularl...

This paper describes an inference system, whose purpose is to determine the cost of evaluating expressions in a strict purely functional language. Upper bounds can be derived for both computation cost and the size of data structures. We outline a static analysis based on this inference system for inferring size and cost information. The analysis is a synthesis of the sized types of Hughes et al., and the polymorphic time system of Dornic et al., which was extended to static dependent costs by Reistad and Gifford. Our main interest in cost information is for scheduling tasks in the parallel execution of functional languages. Using the GranSim parallel simulator, we show that the information provided by our analysis is sufficient to characterise relative task granularities for a simple functional program. This information can be used in the runtime-system of the Glasgow Parallel Haskell compiler to improve dynamic program performance. 1

Analysis of program running time is important for reactive systems, interactive environments, compiler optimizations, performance evaluation, and many other computer applications. It has been extensively studied in many elds of computer science: algorithms [21, 12, 13,40], programming languages [38, 22,31, 35, 34], and systems [36, 29,33,32]. Being able to predict accurate time bounds automatically and e ciently

) David Sands y Department of Computing, Imperial College 180 Queens Gate, London SW7 2BZ email: ds@uk.ac.ic.doc Abstract In this paper we address the technical foundations essential to the aim of providing a semantic basis for the formal treatment of relative efficiency in functional languages. For a general class of "functional" computation systems, we define a family of improvement preorderings which express, in a variety of ways, when one expression is more efficient than another. The main results of this paper build on Howe's study of equality in lazy computation systems, and are concerned with the question of when a given improvement relation is subject to the usual forms of (in)equational reasoning (so that, for example, we can improve an expression by improving any sub-expression). For a general class of computation systems we establish conditions on the operators of the language which guarantee that an improvement relation is a precongruence. In addition, for...

Traditional techniques for designing and analyzing amortized data structures in an imperative setting are of limited use in a functional setting because they apply only to singlethreaded data structures, yet functional data structures can be non-single-threaded. In earlier work, we showed how lazy evaluation supports functional amortized data structures and described a technique (the banker's method) for analyzing such data structures. In this paper, we present a new analysis technique (the physicist's method) and show how one can sometimes derive a worst-case data structure from an amortized data structure by appropriately scheduling the premature execution of delayed components. We use these techniques to develop new implementations of FIFO queues and binomial queues. 1 Introduction Functional programmers have long debated the relative merits of strict versus lazy evaluation. Although lazy evaluation has many benefits [11], strict evaluation is clearly superior in at least one area:...

This paper describes a general approach for optimized live heap space and live heap space-bound analyses for garbage-collected languages.

We definea novel inference system for strictness and totality analysis for the simplytyped lazy lambda-calculus with constants and fixpoints. Strictness information identifies those terms that definitely denote bottom (i.e. do not evaluate to WHNF) whereas totality information identifies those terms that definitely do not denote bottom (i.e. do evaluate to WHNF). The analysis is presented as an annotated type system allowing conjunctions only at "top-level". We give examples of its use and prove the correctness with respect to a natural-style operational semantics. 1 Introduction Strictness analysis has proved useful in the implementation of lazy functional languages as Miranda, Lazy ML and Haskell: when a function is strict it is safe to evaluate its argument before performing the function call. Totality analysis is equally useful but has not be adopted so widely: if the argument to a function is known to terminate then it is safe to evaluate it before performing the function call [1...