List processing in real time on a serial computer
Communications of the ACM
An exercise in proving parallel programs correct
Communications of the ACM
Analysis of an algorithm for real time garbage collection
Communications of the ACM
Multiprocessing compactifying garbage collection
Communications of the ACM
Symbolic Logic and Mechanical Theorem Proving
Symbolic Logic and Mechanical Theorem Proving
On-the-fly garbage collection: an exercise in cooperation
Language Hierarchies and Interfaces, International Summer School
Survey on special purpose computer architectures for AI
ACM SIGART Bulletin
Incremental incrementally compacting garbage collection
SIGPLAN '87 Papers of the Symposium on Interpreters and interpretive techniques
Portable, unobtrusive garbage collection for multiprocessor systems
POPL '94 Proceedings of the 21st ACM SIGPLAN-SIGACT symposium on Principles of programming languages
Distributed copying garbage collection
LFP '86 Proceedings of the 1986 ACM conference on LISP and functional programming
Parallel garbage collection without synchronization overhead
ISCA '85 Proceedings of the 12th annual international symposium on Computer architecture
| Hi-index | 0.00 |
One of the major problems of list processing programs is that of garbage collection. This paper presents a new practical parallel garbage collection algorithm and its improvements, and proposes a special processor for parallel garbage collection. For the parallel garbage collection system, an urgent requirement is to reduce the the garbage collector cycle time that is defined as the total execution time for the marking and reclaiming phase. The effect of improvements discussed here reduces the garbage collector cycle time to one half of that for the original algorithm. The performance of the processor tailored for parallel garbage collection is six times faster than that of an ordinary processor, while it requires a little bit larger amount of hardware than a typical channel controller. This processor satisfies the effectiveness condition for parallelism, even if the list process node consumption rate is high, e.g. when a compiled program is executed.