Fusion, one of the most useful transformation tactics for deriving efficient programs, is the process whereby separate pieces of programs are fused into a single one, leading to an efficient program with no intermediate data structures produced. In this paper, we report our on-going investigation on the design and implementation of an automatic transformation system HYLO which performs fusion transformation in a more systematic and more general way than any other systems. The distinguished point of our system is its calculational feature based on simple application of transformation laws rather than traditional search-based transformation. Keywords Program Transformation, Fusion, Constructive Algorithmics, Bird-Meertens Formalisms 1 INTRODUCTION Program transformation has been advocated as the linchpin of a programming paradigm in which the derivation of efficient programs from naive specification of problems is a formal and mechanically supported process [PP93]. It does not attempt ...

ion versus efficiency For concreteness, let us first examine a number of examples of the type of optimisation that we wish to capture, and the kind of programs on which they operate. This will give us a specific aim when developing the machinery for automating the process, and a yardstick for evaluating our results. 2.1 Minimum depth of a tree Consider the data type of leaf labelled binary trees: dataBtreea = Leaf a j Bin (Btree a)(Btree a) The minimum depth of such a tree is returned by the function mindepth :: Btree a ! Int : mindepth (Leaf a) = 0 mindepth (Bin s t) = min (mindepth s)(mindepth t) + 1 This program is clear, but rather inefficient. It traverses the whole tree, regardless of leaves that may occur at a small depth. A better program would keep track of the `minimum depth so far', and never explore subtrees beyond that current best solution. One possible implementation of that idea is mindepth t = md t 01 md (Leaf a)d m = mindm md (Bin s t)d m = if d 0 m then m...

Dynamic programming is an important algorithm design technique. It is used for solving problems whose solutions involve recursively solving subproblems that share subsubproblems. While a straightforward recursive program solves common subsubproblems repeatedly and often takes exponential time, a dynamic programming algorithm solves every subsubproblem just once, saves the result, reuses it when the subsubproblem is encountered again, and takes polynomial time. This paper describes a systematic method for transforming programs written as straightforward recursions into programs that use dynamic programming. The method extends the original program to cache all possibly computed values, incrementalizes the extended program with respect to an input increment to use and maintain all cached results, prunes out cached results that are not used in the incremental computation, and uses the resulting incremental program to form an optimized new program. Incrementalization statically exploits semantics of both control structures and data structures and maintains as invariants equalities characterizing cached results. The principle underlying incrementalization is general for achieving drastic program speedups. Compared with previous methods that perform memoization or tabulation, the method based on incrementalization is more powerful and systematic. It has been implemented and applied to numerous problems and succeeded on all of them. 1

this paper, we shall formulate accumulations as higher order catamorphisms , and propose several general transformation rules for calculating accumulations (i.e., finding and manipulating accumulations) by calculation-based (rather than a search-based) program transformation methods. Some examples are given for illustration.

Parallel skeletons intend to encourage programmers to build...

In this paper we propose a new method for deriving a practical linear-time algorithm from the specification of a maximum-weight sum problem: From the elements of a data structure x, find a subset which satisfies a certain property p and whose weightsum is maximum. Previously proposed methods for automatically generating linear-time algorithms are theoretically appealing, but the algorithms generated are hardly useful in practice due to a huge constant factor for space and time. The key points of our approach are to express the property p by a recursive boolean function over the structure x rather than a usual logical predicate and to apply program transformation techniques to reduce the constant factor. We present an optimization theorem, give a calculational strategy for applying the theorem, and demonstrate the effectiveness of our approach through several nontrivial examples which would be difficult to deal with when using the methods previously available.

Abstract Many functions have to be written over and over again for different datatypes, either because datatypes change during the development of programs, or because functions with similar functionality are needed on different datatypes. Examples of such functions are pretty printers, pattern matchers, equality functions, unifiers, rewriting functions, etc. Such functions are called polytypic functions. A polytypic function is a function that is defined by induction on the structure of user-defined datatypes. This thesis introduces polytypic functions, shows how to construct and reason about polytypic functions and describes the implementation of the polytypic programming system PolyP. PolyP extends a functional language (a subset of Haskell) with a construct for writing polytypic functions. The extended language type checks definitions of polytypic functions, and infers the types of all other expressions. Programs in the extended language are translated to Haskell.

Correctness-preserving program transformation has recently received a particular attention for compiler optimization in functional programming [Kelsey and Hudak 1989; Appel 1992; Peyton Jones 1996]. By implementing a compiler using many passes, each of which is a transformation for a particular optimization, one can attain a modular compiler. It is no surprise that the modularity would increase if transformations are structured, i.e. constructed in a modular way. Indeed, the program transformation in calculational form (or program calculation) can help us to attain this goal.

The general goal of data mining is to extract interesting correlated information from large

Program fusion techniques have long been proposed as an effective means of improving program performance and of eliminating unnecessary intermediate data structures. This paper proposes a new approach on program fusion that is based entirely on the type signatures of programs. First, for each function, a recursive skeleton is extracted that captures its pattern of recursion. Then, the parametricity theorem of this skeleton is derived, which provides a rule for fusing this function with any function. This method generalizes other approaches that use fixed parametricity theorems to fuse programs. 1 Introduction There is much work recently on using higher-order operators, such as fold [11] and build [8, 5], to automate program fusion [2] and deforestation [13]. Even though these methods do a good job on fusing programs, they are only effective if programs are expressed in terms of these operators. This limits their applicability to conventional functional languages. To ameliorate this pr...