Patent
1992-11-27
1998-10-20
Donaghue, Larry D.
39520073, 39580012, H04L 12407
Patent
active
058260331
ABSTRACT:
A parallel computer and all-to-all communications method. A plurality of processors are connected in an n-dimensional torus network, to provide an optimum communication method and apparatus for completing all-to-all communications within a shortest possible time. In the parallel computer having an n-dimensional rectangular parallelopiped torus network having a.sub.1 .times.a.sub.2 .times. . . . .times.a.sub.n processors, with the maximum value of a.sub.1, a.sub.2, . . . , a.sub.n defined as a.sub.max, it comprises a phase control unit having a phase control table for storing information according to which destination processors are determined for each of the predetermined transmission phases of a.sub.max P/4 (where P indicates the total number of processors) for a one-directional inter-processor connection channel, and of a.sub.max P/8 for a two-directional inter-processor connection channel, and a message transmission unit for transmitting a message to a destination processor listed in the phase control table during the phase in which the present processor is a source processor according to the predetermined phase order.
REFERENCES:
patent: 4645874 (1987-02-01), Fildes
patent: 4933933 (1990-06-01), Dally et al.
patent: 5008882 (1991-04-01), Peterson et al.
patent: 5099496 (1992-03-01), Pope et al.
patent: 5105424 (1992-04-01), Flaig et al.
patent: 5157692 (1992-10-01), Horie et al.
patent: 5163176 (1992-11-01), Flumerfelt et al.
patent: 5218676 (1993-06-01), Ben-Ayed et al.
patent: 5228138 (1993-07-01), Pratt et al.
patent: 5278975 (1994-01-01), Ishihata et al.
Proceedings Supercomputing 88, 14, Nov. 1988, Orlando, USA, pp. 42-47, Pittelli, `Analysis of a 3D Toroidal Network for a Shared Memory Architecture`.
Sixth Annual International Phoenix Conference on Computers and Communications, 25 Feb. 1987, Scottsdale, USA, pp. 14-17, Midkiff, `Architectural Support for Interprocessor Communication in Point-to Point Multiprocessor Networks`.
Computer Architecture News, vol. 19, No. 1, Mar. 1991, New York, U.S., pp. 69-78, Herbordt, `Message-passing Algorithms for SIMD Torus with Coteries`.
Proceedings Supercomputing 89, 13 Nov. 1989, Reno, USA, pp. 436-445, Gross, `Communication in iWarp Systems`.
Linder et al., "An Adaptive and Fault Tolerant Wormhole Routing Strategy for k-ary n-cubes", IEEE Transactions on computers, vol. 40, No. 1, Jan. 1991, pp. 2-12.
Hayashi Ken-ichi
Horie Takeshi
Donaghue Larry D.
Fujitsu Limited
Hollansbee John
LandOfFree
Parallel computer apparatus and method for performing all-to-all does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Parallel computer apparatus and method for performing all-to-all, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Parallel computer apparatus and method for performing all-to-all will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-256434