Run-time code generation (RTCG) would be used routinely if application programmers had a facility with which they could easily create their own run-time code generators, because it would offer benefits both in terms of the efficiency of the code that programmers would produce and the ease of producing it. Such a facility would necessarily have the following properties: it would not require that programmers know assembly language; programmers would have full control over the generated code; the code generator would operate entirely at the binary level. In this paper, we offer arguments and examples supporting these assertions. We briefly describe Jumbo, a system we have built for producing run-time code generators for Java..

An image manipulation system can be thought of as a domain-specific programming language: by composing and manipulating pictures, the user builds up an expression in such a programming language. Each time the picture is displayed, the expression is evaluated. To gain the required efficiency of display, the expression must be optimized and compiled on the fly. This paper introduces a small language that could form the basis of an image manipulation system, and it describes a preliminary implementation. To compile an expression in the language, we inline all function definitions and use algebraic laws of the primitive operations to optimize composites. We apply a form of code motion to recover (as far as possible) the sharing lost in the inlining phase. Finally, we generate intermediate code that is passed to a JIT compiler.

We have implemented a two-stage language, Paradise, for building reusable components which are used to price financial products. Paradise is embedded in Haskell and makes heavy use of type-class based overloading, allowing the second stage to be compiled into a variety of backend platforms. Paradise has enabled us to begin moving away from implementation directly in monolithic Excel spreadsheets and towards a more modular and retargetable approach.