A Self-Routing Permutation Network

نویسندگان

  • David M. Koppelman
  • A. Yavuz Oruç
چکیده

A self–routing permutation network is a connector which can set its own switches to realize any one-to-one mapping of its inputs onto its outputs. Many permutation networks have been reported in the literature, but none with self–routing property, except crossbars and cellular permutation arrays which have excessive cost. This paper describes a self–routing permutation network which has O(log n) bit-level delay and uses O(n log n) bit-level hardware where n is the number of inputs to the network. The network is derived from a complementary Beneš network by replacing each of its two switches in its first stage by what is called a 1–sorter, and recursively defining the switches in the third stage as self– routing networks. The use of 1–sorters results in substantial reduction in both propagation delay and hardware cost when contrasted with O(n) delay and O(n) hardware of the recursively decomposed version of a complementary Beneš network. Furthermore, these complexities match the propagation delay and hardware cost of Batcher’s sorters (only networks, other than crossbars and cellular permutation arrays, which are known to behave like self–routing permutation networks.) More specifically, it is shown that the network of this paper uses about half of the hardware with about four-thirds of the delay of a Batcher’s sorter. † This work is supported in part by the National Science Foundation under grant No:CCR-8708864

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Fast Self-Routing Permutation Switching on an Asymptotically Minimum Cost Network

Permutation switching is a key operation in many computer and communication systems. The well-known BeneS and Waksman permutation networks can be constructed with an asymptotically minimum number of switches, but the best routing algorithms for such networks need O( ( Ig4 n/lg Ig n ) ) time on an n Ig n-processor computer. Other networks that can be used for permutation switching are Batcher’s ...

متن کامل

A virtually nonblocking self - routing permutation network which routes packets in O ( log 2 N ) time

Asymptotically nonblocking networks are O(log2 N) depth self-routing permutation devices in which blocking probability vanishes when N (the number of network inputs) increases. This behavior does not guarantee, also for very large N , that all information always and simultaneously reaches its destination (and consequently that a whole permutation passes through the device) which is a requiremen...

متن کامل

Routing Permutations with Link-Disjoint and Node-Disjoint Paths in a Class of Self-Routable Interconnects

In this paper, we consider efficiently routing permutations in a class of switch-based interconnects. Permutation is an important communication pattern in parallel and distributed computing systems. We present a generic approach to realizing arbitrary permutations in a class of unique-path, self-routable interconnects. It is well-known that this type of interconnect has low hardware cost, but c...

متن کامل

Optimal Self { Routing of Linear - Complement Permutations in Hypercubes 1

Optimal Self{Routing of Linear-Complement Permutations in Hypercubes1 Rajendra Boppana and C. S. Raghavendra Dept. of Electrical Engineering{Systems University of Southern California, Los Angeles, CA 90089{0781 Abstract In this paper we describe an algorithm to route the class of linear-complement permutations on Hypercube SIMD computers. The class of linearcomplement permutations are extremely...

متن کامل

An O(log2 N) Depth Asymptotically Nonblocking Self-Routing Permutation Network

A self-routing multi-logN permutation network is presented and studied. This network has 3 1 0 ~ N 2 depth and N(log~N)(31og~, N -2)/2 nodes, where N is the number of network inputs and 7 a constant very close to 1. A parallel routing algorithm runs in 3 l o m 2 time on thii network. The overall system (network and algorithm) can work in pipeline and it is asymptotically nonblocking in the sens...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:
  • J. Parallel Distrib. Comput.

دوره 10  شماره 

صفحات  -

تاریخ انتشار 1989