Abstract
This paper presents a cost-effective scheme for partitioning large data flow graphs. Standard data flow machine architectures are assumed in this work. The objective is to reduce the overhead due to token transfers through the communication network of the machine. When this scheme is employed on large graphs, the load distribution on the rings of the data flow machine is also improved. A canonical form of a data flow graph is introduced to establish the relationship between the communication overhead and the size reduction of the partition cut-set. General lower estimates on the overhead are derived in terms of processing and transmission delay parameters of the machine. The method uses heuristics and an evaluation function to guide the partition algorithm. Some implications of the proposed method on the organization of the data flow machines are discussed.
- 1 A.L. Davis and R.M. Keller, " Data Flow Program Graphs," Computer, Vo1.15, No.,?, Feb. 1982, pp. 26-41Google Scholar
Digital Library
- 2 J.B. Dennis, "Data Flow Supercomputers," Computer, Vol.13, No.11, Nov. 1980, pp. 48-56.Google Scholar
Digital Library
- 3 Arvind et. al., "The Tagged Token Dataflow Architecture", MIT, Aug 1983.Google Scholar
- 4 Arvind & R. Iannucci, "Two Fundamental Issues in Multiprocessing - The Data Flow Solution," Computation Structures Group Memo 226-2, MIT, Jul. 1983Google Scholar
- 5 I. Watson and J. Gurd, "A practical Data Flow Computer," Computer, Vol.15 , No.2, Feb. 1982, pp. 51-57Google Scholar
Digital Library
- 6 I. Watson and J. Gurd, "A Prototype Data Flow Computer with Token Labeling," AFIPS Conf. Proc. Nat'l Comput. Conf. June 1979.Google Scholar
- 7 Arvind and K.P. Gostelow, "The U-Interpreter," Computer, Vo1.15, No.4 Feb. 1982.Google Scholar
- 8 Arvind, "Decomposing a Program for Multiple PIV cessor Systems," IEEE March 1980Google Scholar
- 9 J.R. Gurd, C.C. Kirkham and I. Watson, " The Machester Prototype Data Flow Computer", CACM, Vol.28, No. 1, Jan 1985, pp.34-52. Google Scholar
Digital Library
- 10 D. D. Gajski et. al., "A Second Opinion on Data Flow Machines and Languages," Computer, Vol.15, No. 2, Feb 1982.Google Scholar
- 11 T.L. Chang and P.D. Fisher, "A Block-Driven Data Flow Processor," IEEE Parallel Processing Conf. 1981.Google Scholar
- 12 A.Aho, Hopcroft and J. Uilman, The Design and Analysis of Computer Algorithms, Addison-Wesley, 1974. Google Scholar
Digital Library
- 13 A. Tanenbaum, Computer Networks, Prentice Hall, 1981. Google Scholar
Digital Library
- 14 Y.N. Patt, W-M. Hwu and M. Shebanow, "HPS, a new microarchitecture: rationale and introduction," 18th IEEE Microprogramming Workshop, December 1985. Google Scholar
Digital Library
- 15 D. Gajski, D. Kuck, Lawrie and D. Sameh, "CEDAR- a large scale multiprocessor," Computer Architecture News, March 1983. Google Scholar
Digital Library
Index Terms
Data flow graph partitioning to reduce communication cost
Recommendations
Data flow graph partitioning to reduce communication cost
MICRO 19: Proceedings of the 19th annual workshop on MicroprogrammingThis paper presents a cost-effective scheme for partitioning large data flow graphs. Standard data flow machine architectures are assumed in this work. The objective is to reduce the overhead due to token transfers through the communication network of ...
Balanced Connected Graph Partition
Algorithms and Discrete Applied MathematicsAbstractWe study a variation of the graph partition problem on colored graphs called the [inline-graphic not available: see fulltext] problem. We are given a connected non-unicolor graph G where the vertices in G have colored either [inline-graphic not ...
Partition the vertices of a graph into induced matchings
The induced matching partition number of a graph G, denoted by imp(G), is the minimum integer k such that V(G) has a k-partition (V1, V2,..., Vk) such that, for each i, 1 ≤ i ≤ k, G[Vi], the subgraph of G induced by Vi, is a 1-regular graph. This is ...






Comments