Effects for Efficiency: Asymptotic Speedup with First-Class Control

TL;DR

This paper demonstrates that effect handlers in programming languages can asymptotically improve runtime efficiency for certain functions, even with mutable state, marking a novel result for control operators.

Contribution

It proves that effect handlers enable asymptotic speedups in runtime complexity for a class of functions, a first for control operators.

Findings

Effect handlers improve asymptotic runtime complexity.

Efficiency gap persists with mutable state.

First formal proof of control operators' efficiency benefits.

Abstract

We study the fundamental efficiency of delimited control. Specifically, we show that effect handlers enable an asymptotic improvement in runtime complexity for a certain class of functions. We consider the generic count problem using a pure PCF-like base language $\lambda_b$ and its extension with effect handlers $\lambda_h$. We show that $\lambda_h$ admits an asymptotically more efficient implementation of generic count than any $\lambda_b$ implementation. We also show that this efficiency gap remains when $\lambda_b$ is extended with mutable state. To our knowledge this result is the first of its kind for control operators.