A technique for overlapping computation and communication for block recursive algorithms
β Scribed by GUPTA, S. K. S.; HUANG, C.-H.; SADAYAPPAN, P.; JOHNSON, R. W.
- Publisher
- John Wiley and Sons
- Year
- 1998
- Tongue
- English
- Weight
- 175 KB
- Volume
- 10
- Category
- Article
- ISSN
- 1040-3108
No coin nor oath required. For personal study only.
β¦ Synopsis
This paper presents a design methodology for developing efficient distributed-memory parallel programs for block recursive algorithms such as the fast Fourier transform (FFT) and bitonic sort. This design methodology is specifically suited for most modern supercomputers having a distributed-memory architecture with a circuit-switched or wormhole routed mesh or a hypercube interconnection network. A mathematical framework based on the tensor product and other matrix operations is used for representing algorithms. Communication-efficient implementations with effectively overlapped computation and communication are achieved by manipulating the mathematical representation using the tensor product algebra. Performance results for FFT programs on the Intel Paragon are presented.
π SIMILAR VOLUMES
This paper gives a brief and systematic presentation of the computer simulation techniques of satellite communications, especially for a mobile satellite channel, based on a summarization of the author's exploration and experience in this area. It includes the equivalent complex baseband notation (E
A technique for mpidly computing the capacitances of two-dimensional conducting structures is presented in fhis article. The method relies on a sparse multiresolution representation of the interactions between subsections of the conductors. This decomposition is used to substantially reduce the comp
Three domain decomposition methods for saddle point problems are introduced and compared. The first two are blockdiagonal and block-triangular preconditioners with diagonal blocks approximated by an overlapping Schwarz technique with positive definite local and coarse problems. The third is an overl