An orthogonal multiprocessor for parallel scientific computations
A new multiprocessor architecture, called orthogonal multiprocessor (OMP), is proposed in this paper. This OMP architecture has a simplified bussing structure and partially shared memory, which compares very favorably over fully shared-memory multiprocessors using crossbar switch, multiple buses, or multistage networks. The higher performance comes mainly from significantly increased memory bandwidth, fully exploited parallelism, reduced communication overhead, and lower hardware control complexities. Parallel algorithms being mapped include matrix arithmetic, linear system solver, FFT, array sorting, linear programming, and parallel PDE solutions. In most cases, linear speedup can be achieved on the OMP system. The OMP architecture provides linearly scalable performance and is well suited for building special-purpose scientific computers such as for signal/image processing, machine sorting, linear system solvers, and PDE machines, etc.
- Research Organization:
- Dept. of Electrical Engineering-Systems, Univ. of Southern California, Los Angeles, CA (US); Dept. of Elecrtrical and Computer Engineering, Carnegie-Mellon Univ., Pittsburgh, PA (US)
- OSTI ID:
- 6024472
- Journal Information:
- IEEE Trans. Comput.; (United States), Journal Name: IEEE Trans. Comput.; (United States) Vol. 38:1; ISSN ITCOB
- Country of Publication:
- United States
- Language:
- English
Similar Records
Probabilistic reliability analysis of multiprocessor architectures
Models for parallel computation on distributed-memory multiprocessors
Related Subjects
990210* -- Supercomputers-- (1987-1989)
ALGORITHMS
ARRAY PROCESSORS
COMMUNICATIONS
COMPUTER ARCHITECTURE
COMPUTER CODES
COMPUTER NETWORKS
IMAGE PROCESSING
LINEAR PROGRAMMING
MATHEMATICAL LOGIC
ORTHOGONAL TRANSFORMATIONS
PARALLEL PROCESSING
PROCESSING
PROGRAMMING
TRANSFORMATIONS