On the Specialization of Online Program Specializers

Program specializers improve the speed of programs by performing some of the programs' reductions at specialization time rather than at runtime. This specialization process can be time-consuming; one common technique for improving the speed of the specialization of a particular program is to specialize the specializer itself on that program, creating a custom specializer, or program generator, for that particular program. Much research has been devoted to the problem of generating e cient program generators, which do not perform reductions at program generation time which could instead have been performed when the program generator was constructed. The conventional wisdom holds that only o ine program specializers, which use binding time annotations, can be specialized into such e cient program generators. This paper argues that this is not the case, and demonstrates that the specialization of a nontrivial online program specializer similar to the original \naive MIX" can indeed yield an e cient program generator. The key to our argument is that, while the use of binding time information at program generator generation time is necessary for the construction of an e cient custom specializer, the use of explicit binding time approximation techniques is not. This allows us to distinguish the problem at hand (i.e., the use of binding time information during program generator generation) from particular solutions to that problem (i.e., o ine specialization). We show that, given a careful choice of specializer data structures, and su ciently powerful specialization techniques, binding time information can be inferred and utilized without the use of explicit binding time approximation techniques. This allows the construction of e cient, optimizing program generators from online program specializers.