Synchronization and Arbitration in Digital Systems

11

Arbitration

11.1 INTRODUCTION

Arbitration circuits, or simply arbiters, are commonplace in digital systems where a restricted number of resources are allocated to different user or client processes. Typical examples are systems with shared busses, multi-port memories, and packet routers, to name but a few. The problem of arbitration between multiple clients is particularly difficult to solve when it is posed in asynchronous context, where the time frames in which all clients operate may be different and vary with time. Here, the decision which resource is to be given to which process must be made not only on the basis of some data values, such as priority codes or ‘geographical’ positions of the clients, but also depending on the relative temporal order of the arrival of requests from the clients. In a truly asynchronous scenario, where no absolute clocking of events is possible, the relative timing order cannot be determined by any means other than making the arbiter itself sense the request signals from the clients at the arbiter's inputs. This implies the necessity to use devices such as synchronizers (Sync) and mutual exclusion (MUTEX) elements, described in previous chapters. Such devices have the required sensing ability because they operate at the analog level and they could be placed at the front end of the arbiter.

Why does the problem of arbitration need to be considered in asynchronous context? Can we build a clocked arbiter, in which the clock would tell the rest ...