Pipelined data-parallel algorithms. Part 1; Concept and modeling
- Dept. of Computer and Information Science, New Jersey Institute of Technology, Newark, NJ (US)
- Michigan State Univ., East Lansing, MI (USA). Dept. of Computer Science
As parallel processors are becoming popular, the need for designing efficient parallel algorithms becomes imminent. Pipelined data-parallel algorithms are a class of algorithms which use pipelined operations and data level partitioning to achieve parallelism. Applications which involve data parallelism and recurrence relations are good candidates for this kind of algorithm. Pipelined data-parallel computations are ideal for distributed-memory multicomputers. By controlling the granularity through data partitioning and overlapping the operations through pipelining, it is possible to achieve a balanced computation on multicomputers. In this paper, the basic concept of pipelined data-parallel algorithms is introduced by first contrasting with other styles of computations and then by a single example, a pipelined image distance transformation algorithm. The authors present an analytic model for modeling pipelined data-parallel computation on multicomputers. The model uses timed Petri nets to describe data pipelining operations. As a case study, the model is applied to a pipelined matrix multiplication algorithm. Predicted results match closely with the measured performance on a 64-node NCUBE.
- OSTI ID:
- 6356451
- Journal Information:
- IEEE Transactions on Parallel and Distributed Systems; (USA), Vol. 1:4; ISSN 1045-9219
- Country of Publication:
- United States
- Language:
- English
Similar Records
Pipelined data-parallel algorithms. Part 2; Design
Large-grain pipelining on hypercube multiprocessors
Related Subjects
HYPERCUBE COMPUTERS
DESIGN
PARALLEL PROCESSING
ALGORITHMS
ARRAY PROCESSORS
COMPUTERIZED SIMULATION
DISTRIBUTED DATA PROCESSING
PERFORMANCE TESTING
COMPUTERS
DATA PROCESSING
MATHEMATICAL LOGIC
PROCESSING
PROGRAMMING
SIMULATION
TESTING
990200* - Mathematics & Computers