Parallel graph contraction
The parallel-tree contraction technique of Miller and Reif has proved to be a valuable tool in many parallel graph algorithms. This paper provides a similar contraction technique that applies not only to trees, but also to general graphs. It also provides a second, potentially faster contraction technique for bounded-degree graphs (and general graphs when an embedding is known). The paper, uses the technique of parallel contraction in a more-general setting. It is shown that a remarkably simple contraction algorithm that uses (n + e)/lg n processors reduces a connected n-node e-edge graph to a single node in O(lg squared n) steps and a second simple algorithm which uses the first as a subroutine.
- Research Organization:
- Massachusetts Inst. of Tech., Cambridge, MA (USA). Microsystems Research Center
- OSTI ID:
- 5145457
- Report Number(s):
- AD-A-211916/2/XAB; VLSI-MEMO--89-535
- Country of Publication:
- United States
- Language:
- English
Similar Records
Communication-efficient parallel graph algorithms. Technical report
Communication-efficient parallel-graph algorithms. Master's thesis