Abstract:Warm-up transformation is an important preprocess for shortcut fusion. In this paper, we formalize the warm-up transformation by proposing several general and powerful calculation rules that can be directly implemented with help of higher-order pattern matching. The newly proposed warm-up transformation cannot only deal with pro- grams that existing methods may fail, but also suitable for efficient implementation. Constructing programs from components, i.e., modules, is an important technique in designing large software. Dividing problems into small modules realizes local- ization of problems. Consequently, programs are easy to maintain. If each module is maintained as a versatile component, reusability is improved. The problem is, however, in its inefficiency due to unnecessary intermediate data structures passed between modules. Improvement of performance can be achieved if these interme- diate data structures are removed by automatic analysis and automatic transforma- tion. Fusion (a.k.a. Deforestation) (Wad90) is a technique of elimination of inter- mediate data structures. Descendent from it, Gill et al. (GLP93) proposes short- cut fusion, which is a single, local transformation rule, to ease the implementa- tion (PTH01). The precondition of shortcut fusion assumes the producer function to be expressed in terms of build. If we could prepare all the functions in terms of foldr and build, shortcut fusion could remove intermediate data structures automatically. Functional programmer usually describes program as recursive definitions, since foldr form and build form are less natural than recursive definitions. Therefore, the problem is how to derive foldr form and build form from a recursive definition. There is an approach to au- tomatically derive the foldr form (SF93, HIT96). Therefore, the problem is settled down into the derivation of build form from a recursive function. Warm fusion (LS95) accompanies shortcut fusion with a preprocess (hereafter we call it warm-up transformation). It transforms recursive functions into func- tions in terms of build automatically. The technique has been implemented with the Stratego language (JV00), which is based on term rewriting framework. The

A Modular Derivation Strategy via Fusion and Tupling

Towards a Modular Program Derivation via Fusion and Tupling

A Case Study on a Modular Transformation Strategy

Parallelization in Calculational Forms.

Deriving Parallel Codes Via Invariants

Tupling Calculation Eliminates Multiple Data Traversals.

Formal Derivation of Efficient Parallel Programs by Construction of List Homomorphisms.

A Simultaneous Strategy For Dynamic Optimization Based On Symbolic Derivation

Formal Derivation of Parallel Program for 2-Dimensional Maximum Segment Sum Problem

Superfusion: Eliminating Intermediate Data Structures Via Inductive Synthesis

Modular quantum signal processing in many variables

Fusing Gathers with Integer Linear Programming

Memory fartitioning-based modulo scheduling for high-level synthesis.

Pearl: A Multi-Derivation Approach to Efficient CFL-Reachability Solving

Domain-Specific Optimization Strategy For Skeleton Programs

A Two-Stage Pipelined Architecture for Parallel Modular Exponentiation

A Modular Multiplier Implemented with Truncated Multiplication

MoTCoder: Elevating Large Language Models with Modular of Thought for Challenging Programming Tasks

Program optimizations and transformations in calculation form

Calculation Rules for Warming-up in Fusion Transformation

Modular Monadic Program Slicing.