Efficient multiple and predicated dispatching

Abstract

The speed of message dispatching is an important issue in the overall performance of object-oriented programs. We have developed an algorithm for constructing efficient dispatch functions that combines novel algorithms for efficient single dispatching, multiple dispatching, and predicate dispatching. Our algorithm first reduces methods written in the general predicate dispatching model (which generalizes single dispatching, multiple dispatching, predicate classes and classifiers, and pattern-matching) into ones written using a simpler multimethod dispatching model. Our algorithm then computes a strategy for implementing multiple dispatching in terms of sequences of single dispatches, representing the strategy as a lookup DAG. Finally, our algorithm computes an implementation strategy separately for each of the single dispatches, producing for each dispatch a dispatch tree, which is a binary decision tree blending class identity tests, class range tests, and table lookups. Our algorithm exploits any available static information (from type declarations or class analysis) to prune unreachable paths Tom the lookup DAG, and uses any available dynamic profile information to minimize the expected time to search the dispatch trees. We measure the effectiveness of our dispatching algorithms on a collection of large Cecil programs, compiled by the Vortex optimizing compiler, showing improvements of up to 30% over already heavily optimized baseline versions.

Redundancy-free residual dispatch

“…In this setting, residual dispatch at a join point shadow can be viewed as the evaluation of a finitary Boolean function f φ : B n → B, where B = {0, 1}, the so-called dispatch function [9]; whether an advice is applicable depends solely on the prior evaluation of the n atomic pointcuts occurring in φ, the residue in question. Residual dispatch thus bears close resemblance to predicate dispatch [12], even though the function's range is restricted to two outcomes: an advice is either applicable or not.…”

“…Then the sinks are relabeled according to g. Finally, subsequent applications of the reduction rules (cf. Figure 5) yield the reduced OBDD representing h. Decision diagrams have been employed, under the name of lookup DAGs, for the efficient implementation of both multiple and predicate dispatch [9]. But these decision diagrams are not necessarily binary; this complicates synthesis.…”

Classification via two-way comparisons

“…In this case a 2wcdt can be significantly smaller than a lookup table, and, likewise, faster and smaller than a conventional search tree, because a search tree has to identify a given query q (or the inter-key interval that q lies in) whereas a decision tree only has to identify q's class. Because they can be faster and more compact, 2wcdts are used in applications such as dispatch trees, which allow compilers and interpreters to quickly resolve method implementations for objects declared with type inheritance [2,3]. (Each type is assigned a numeric ID via a depth-first search of the inheritance digraph.…”

“…Chambers and Chen give a heuristic to construct low-cost 2wcdts, but leave open whether minimum-cost 2wcdts can be found in polynomial time [2,3]. We give the first polynomial-time algorithm to find minimum-cost 2wcdts.…”