Mark scan garbage collection on a distributed architecture
MH Derbyshire - Lisp and Symbolic Computation, 1990 - Springer
MH Derbyshire
Lisp and Symbolic Computation, 1990•SpringerIn any computer architecture designed for the evaluation of declarative languages, efficient
reclamation of redundant (garbage) storage is essential. High performance, exploiting the
inherent parallelism of such languages, is now typically achieved by parallel architectures,
computational graphs thus being distributed across many processing elements. This paper
describes a real-time mark-scan garbage collection algorithm for a distributed machine with
local store. The algorithm allows collection on a per-process basis, and several collections …
reclamation of redundant (garbage) storage is essential. High performance, exploiting the
inherent parallelism of such languages, is now typically achieved by parallel architectures,
computational graphs thus being distributed across many processing elements. This paper
describes a real-time mark-scan garbage collection algorithm for a distributed machine with
local store. The algorithm allows collection on a per-process basis, and several collections …
Abstract
In any computer architecture designed for the evaluation of declarative languages, efficient reclamation of redundant (garbage) storage is essential. High performance, exploiting the inherent parallelism of such languages, is now typically achieved by parallel architectures, computational graphs thus being distributed across many processing elements.
This paper describes a real-time mark-scan garbage collection algorithm for a distributed machine with local store. The algorithm allows collection on a per-process basis, and several collections may run concurrently and asynchronously. Both non real-time and real-time versions are provided.
It is hoped to publish implementation results for the algorithm when such are available.
Springer
