S. Lillevik, Intel Corporation. Touchstone program overview. In Proceedings of the Fifth Distributed Memory Computers Conference, 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....routing [15] the ideas apply to virtual cut through [21] and store and forward networks as well. Overloaded Channels Figure 1: Four packets and their routing paths under deterministic, dimension order routing. 2 The Problem Most existing multicomputer routing networks use deterministic routing [32, 30, 13, 31]. Although there are numerous paths between any source and destination, in order to avoid deadlock, deterministic routing defines a single path from source to destination. Fixed, single path routing prevents effective use of the network s density of physical interconnection because the physical ....

S. Lillevik, Intel Corporation. Touchstone program overview. In Proceedings of the Fifth Distributed Memory Computers Conference, 1990.

....memory is distributed across the processing nodes; only the data in local memory can be accessed directly. Access to data in remote memories is supported by message passing between processors. Direct networks, represented by grid or mesh networks, have been used predominantly in multicomputers [47, 45, 23, 46]. However, as direct networks gain acceptance in shared memory multiprocessor designs, distinguishing these machines by network topology is less appropriate. Though indirect networks provide several advantages such as the topological equidistance property, they suffer from a significant drawback: ....

....of the parallelism exploited and the communication reducing techniques used. Hence, throughout the thesis, we refer to them as multicomputer networks implying communication networks for both types of concurrent computers. 1. 2 The Problem Most multicomputers use dimension order routing [47, 45, 23, 46] due to its simplicity and deadlock freedom. Under dimension order routing, a message is routed successively in each of the dimensions. Since each message passes through the network dimensions in the same order, interdimensional cycles causing deadlock cannot form. Within each dimension, routing ....

[Article contains additional citation context not shown here]

S. Lillevik, Intel Corporation. Touchstone Program Overview. In Proceedings of the Fifth Distributed Memory Computers Conference, 1990.

....been touted as scalable parallel architectures, in fact their scalability is limited by the performance of their interconnection networks. One reason why networks do not achieve their full potential bandwidth is restrictive routing policies. Most existing multicomputers use deterministic routing [13, 11, 6, 12] due to its simplicity. Any deter 1 The research described in this paper was supported in part by National Science Foundation grant CCR 9209336, Office of Naval Research grant N00014 92 J 1961, and National Aeronautics and Space Administration grant NAG 1 613. Additional support has been ....

S. Lillevik, Intel Corporation. Touchstone Program Overview. In Proceedings of the Fifth Distributed Memory Computers Conference, 1990.

....iPSC 2 [2] increased by a factor of four. More importantly, they provided lower latency, circuit switched communication networks. The improvement in both computation and communication performance is even more striking in the coming generation of multicomputers (e.g. the Intel Touchstone system [23]) Node computation speed is expected to increase by an additional order of magnitude and communication performance should increase by two orders of magnitude. In contrast to revolutionary hardware changes, multicomputer software has evolved slowly. Traditional programming languages such as C and ....

Lillevik, S. L. Touchstone Program Overview. In Proceedings of the 5 th Distributed Memory Computing Conference (April 1990), Association For Computing Machinery, pp. 647--657.

....(virtual cut through or wormhole routing) the ideas apply to store and forward networks as well. 2 The Problem Overloaded Channels Figure 1: Four packets and their routing paths under deterministic, dimension order routing. Most existing multicomputer routing networks use deterministic routing [22, 20, 8, 21]. Although there are numerous paths between any source and destination, in order to avoid deadlock, deterministic routing defines a single path from source to destination. This means that the interconnection networks cannot make effective use of the density of their physical interconnection ....

S. Lillevik, Intel Corporation. Touchstone program overview. In Proceedings of the Fifth Distributed Memory Computers Conference, 1990.

....teraflops within three years. All these systems are constructed by coupling large numbers of standard, high performance microprocessors via a high speed communication network. For example, the Paragon XP S parallel system [17] the commercial realization of the Intel DARPA Touchstone project [19], is based on the high performance i860 XP microprocessor and provides 200 megabyte second, full duplex communication links with software message passing latency measured in microseconds; the peak performance of the largest hardware configuration approaches 300 gigaflops. As massively parallel, ....

Lillevik, S. L. Touchstone Program Overview. In Proceedings of the Fifth Distributed Memory Computing Conference (1990), IEEE Computer Society, pp. 647--657.

....effectively emulates a fully connected network. See [7] and [22] for more details on the iPSC 860 architecture. The iPSC 860 is also known as the Touchstone Gamma prototype, since it represents an early phase of Intel s Touchstone project. Future Touchstone prototypes will scale up to 2048 nodes [25], and are candidate architectures for the next generation of global climate models. The iPSC 860 at Oak Ridge National Laboratory, which was used for our performance measurements, has 128 nodes, each of which is a 40 MHz i860 processor with a peak execution rate of 80 Mflops for 32 bit floating ....

S. L. Lillevik, Touchstone program overview, in The Fifth Distributed Memory Computing Conference, D. W. Walker and Q. F. Stout, eds., IEEE Computer Society Press, Los Alamitos, CA, 1990, pp. 647--657.

....a set S is jfy 2 S j y xgj. At the end of the algorithm the selected element must be at the center processor of the mesh. Because of the simplicity of its architecture, the mesh model is a well studied and popular model of parallel computation [12] and is the basis for several parallel machines [1, 14]. We consider a standard model [6, 7, 12, 13, 16] in which a processor is allowed to communicate one packet of information to each of its neighbors during a single time step and to queue a constant number of packets between time steps. Any selection algorithm requires n Gamma 1 steps, since this ....

S. L. Lillevik. Touchstone program overview. In Proceedings of the 5th Distributed Memory Computing Conference, Charleston, SC, April 9-12 1990.

....NASA Contract Number NAG 1 613, and by grants from Apple Computer and the Digital Equipment Corporation External Research Program. projected emergence of massively parallel MIMD systems (e.g. the Intel Touchstone project goals include a 2048 node system based on i860 microprocessor technology [7]) understanding and correcting application and system performance bottlenecks become crucial. Although it is now architecturally feasible to construct scalable parallel systems that contain thousands of processors, there is no discipline or set of standard software libraries for evaluating the ....

Lillevik, S. L. Touchstone Program Overview. In Proceedings of the Fifth Distributed Memory

....switches (one port for positive direction, the negative direction and the processor element) in each dimension. Router latency is approximately 50ns and channel data rates are as high as 90MB s, using byte wide links. Derivatives of MRCs are used in several research Cost of Adaptivity 24 machines [1, 20, 30]. The J Machine Router The J Machine is a fine grained concurrent computer developed at MIT [16, 29] The J machine network is a three dimensional mesh, with bidirectional 9 bit channels, and dimension order, wormhole routing. The J Machine network uses two virtual channels to support two ....

S. Lillevik, Intel Corporation. Touchstone program overview. In Proceedings of the Fifth Distributed Memory Computers Conference, 1990.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC