Interconnection Networks and Mapping and Scheduling Parallel Computations: DIMACS Workshop, February 7-9, 1994

Pirmais vāks
Derbiau Frank Hsu, Arnold L. Rosenberg, Dominique Sotteau
American Mathematical Soc., 1995. gada 1. janv. - 342 lappuses
This book contains the refereed proceedings of a DIMACS Workshop on Massively Parallel Computation.

No grāmatas satura

Atlasītās lappuses

Saturs

Dense bus networks of diameter
9
On broadcasting schemes in restricted optical passive star systems
19
Restricted routing and wide diameter of the cycle prefix network
31
Permutation routing via Cayley graphs with an example for bus intercon
47
Using helpful sets to improve graph bisections
57
Modification of consecutived digraphs
75
Highly adaptive wormhole routing algorithms for Ndimensional torus
87
Conflictfree access to constantperimeter rectangular subarrays
105
Communications in optically interconnected parallel computer systems
181
Faulttolerant Kautz networks
201
Asynchronous packet routers
211
Cayley digraphs of finite cyclic groups with minimal average distance
229
Shuffled tree based faulttolerant hierarchical interconnection networks
251
Restricted connectivity and restricted fault diameter of some interconnec
267
Sorting and selection on interconnection networks
275
Towards a simple construction method for Hamiltonian decomposition
297

Makespan minimization of task graphs with random task running times
125
Scheduling of structured and unstructured computation
139
The problem of contention
173
Generalized reduced hypercube interconnection networks for massively par
307
List of Participants
327
Autortiesības

Citi izdevumi - Skatīt visu

Bieži izmantoti vārdi un frāzes

Populāri fragmenti

290. lappuse - GE Blelloch, CE Leiserson, BM Maggs, CG Plaxton, SJ Smith, and M. Zagha. A comparison of sorting algorithms for the Connection Machine CM2.
324. lappuse - A survey of wormhole routing techniques in direct networks,
87. lappuse - The flit at the head of a message governs the route. As the header flit advances along the specified route, the remaining flits follow it in a pipeline fashion. If the header encounters a channel already in use, it is blocked until the channel is freed; the flow control within the network blocks...
88. lappuse - However, deadlocks may appear if the routing algorithms are not carefully designed. A deadlock in the interconnection network of a multicomputer occurs when no message can advance toward its destination because the queues of the message system are full. The size of the queues strongly influences the probability to reach a deadlocked configuration.
325. lappuse - Department of Electrical and Computer Engineering New Jersey Institute of Technology Newark. NJ 07102. USA GRAHAM C GOODWIN Department of Electrical and Computer Engineering University of Newcastle New South Wales.
200. lappuse - Department of Computer Science Hong Kong University of Science and Technology Clear Water Bay, Kowloon, Hong Kong Email: {zhangfan, chanson} @cs.ust.hk Abstract Multimedia applications over the Internet are becoming increasingly popular.
88. lappuse - These algorithms are based on a structured buffer pool. However, with wormhole routing, buffer allocation cannot be restricted, because flits have no routing information. Once the header of a message has been accepted by a channel, the remaining flits must be accepted before the flits of any other message can be accepted. So, routing must be restricted to avoid deadlock.
277. lappuse - Therefore, there are exactly n edges going out of (and coming into) any vertex. If a hypercube processor can handle only one edge at any time step, this version of the hypercube will be called the sequential model. Handling (or processing) an edge here means either sending or receiving a key along that edge. A hypercube model where each processor can process all its incoming and outgoing edges in a unit step is called the parallel model [9].
88. lappuse - Dally [9] has proposed a methodology to design static routing algorithms under general assumptions. He defines a channel dependency graph and establishes a total order among channels. Routing is restricted to visit channels in decreasing or increasing order to eliminate cycles in the channel dependency graph. This methodology has been applied to the design of routing chips for multicomputers [8] and multicomputer nodes with integrated communication support [2].
10. lappuse - The distance between two vertices u and v is the length of a shortest path between them. The diameter of H is the maximum of the distances over all pairs of vertices, and is denoted by D{H).

Bibliogrāfiskā informācija