Tag Elimination and Jones-Optimality

Abstract

Abstract. Tag elimination is a program transformation for removing unnecessary tagging and untagging operations from automatically generated programs. Tag elimination was recently proposed as having immediate applications in implementations of domain specific languages (where it can give a two-fold speedup), and may provide a solution to the long standing problem of Jones-optimal specialization in the typed setting. This paper explains in more detail the role of tag elimination in the implementation of domain-specific languages, presents a number of significant simplifications and a high-level, higher-order, typed self-applicable interpreter. We show how tag elimination achieves Jones-optimality.

Jones optimality, binding-time improvements, and the strength of program specializers

“…The results in this paper extend previous work on Jones optimality [18,28,22,32] and the interpretive approach [11,13,34]. We will see that Jones optimality [18] plays a key role in the power of binding-time improvers and, together with static expression reduction, establishes a certain kind of non-triviality.…”

“…For other partial evaluators, such as FCL-mix [15], it is impossible to write a self-interpreter that makes them Jones optimal. Recent work [32] on Jones optimality concerns tag-elimination when specializing self-interpreters for strongly typed languages.…”

Finally tagless, partially evaluated: Tagless staged interpreters for simpler typed languages

“…In other words, the algebraic data types above fail to express in the metalanguage that the object program is well-typed. This failure necessitates tagging and nonexhaustive pattern-matching operations that incur a performance penalty in interpretation and impair optimality in partial evaluation (Jones et al 1993;Taha et al 2001). In short, the universal-type solution is unsatisfactory because it does not preserve the type of the encoded term.…”

“…For example, Taha et al (2001) say that "expressing such an interpreter in a statically typed programming language is a rather subtle matter. In fact, it is only recently that some work on programming type-indexed values in ML (Yang 2004) has given a hint of how such a function can be expressed."…”

“…Our initial motivation came from several papers that justify advanced type systems, in particular GADTs, by embedded interpreters (Pašalić et al 2002;Peyton Jones et al 2006;Taha et al 2001;Xi et al 2003) and CPS transformations Guillemette and Monnier 2006;Shao et al 2005). We admire all this technical machinery, but these motivating examples do not need it.…”