Bounding Cache Miss Costs of Multithreaded Computations Under General Schedulers

TL;DR

This paper provides bounds on cache miss costs for multithreaded algorithms under general schedulers, extending known results for randomized work stealing to broader scheduling scenarios.

Contribution

It introduces bounds that match or improve existing cache miss cost estimates for multithreaded algorithms under arbitrary schedulers.

Findings

Bounds match or improve upon $O(Q+S imes (M/B))$ for randomized work stealing

Applicable to arbitrary schedulers, not just RWS

Enhances understanding of cache costs in multithreaded scheduling

Abstract

We analyze the caching overhead incurred by a class of multithreaded algorithms when scheduled by an arbitrary scheduler. We obtain bounds that match or improve upon the well-known $O(Q+S \cdot (M/B))$ caching cost for the randomized work stealing (RWS) scheduler, where $S$ is the number of steals, $Q$ is the sequential caching cost, and $M$ and $B$ are the cache size and block (or cache line) size respectively.