Summary: Real­time Concurrent Collection
on Stock Multiprocessors
Andrew W. Appel 1
John R. Ellis 2
Kai Li 1
Abstract
We have designed and implemented a copying garbage­collection algorithm that is
efficient, real­time, concurrent, runs on commerial uniprocessors and shared­memory
multiprocessors, and requires no change to compilers. The algorithm uses standard
virtual­memory hardware to detect references to ``from space'' objects and to syn­
chronize the collector and mutator threads. We have implemented and measured a
prototype running on SRC's 5­processor Firefly. It will be straightforward to merge
our techniques with generational collection. An incremental, non­concurrent version
could be implemented easily on many versions of Unix.
Introduction
This paper presents the first copying garbage­collection algorithm that is efficient,
real­time, concurrent, runs on stock commercial uniprocessors and multiprocessors,
and requires no change to compilers.
A collection algorithm is efficient if the amortized cost to allocate, access, and
collect an object is small compared to the cost of initializing the object.

Source: Appel, Andrew W. - Department of Computer Science, Princeton University

Collections: Computer Technologies and Information Sciences