previous | contents | next


Digital technology continues to change and with it changes both our understanding of digital systems and the nature of the cost-effective ways to create them. The development has continued to be towards greater performance and the accomplishment of more complex information processing tasks. Simultaneously it has continued toward cheaper, more reliable systems, and the accomplishment .of an ever widening spectrum of tasks. This joint development gives no hint yet of slackening.

One particular item in this development is the current emergence of the register-transfer level (RT level) to the status of a full fledged system level. The conceptualization of digital systems in terms of registers and data transfers goes back to the beginnings of digital computers. But in practice logical design has been carried out in terms of combinational and sequential circuits, with the concepts of registers, functional units and transfer paths playing only a heuristic role. Examination of any of the standard books on logical design will confirm this, as will the degradation of the formal concern with the register-transfer level since the work of Bartee, Lebow and Reed (1962) which provided a high water mark in the explicit treatment of this level.

The current maturation of the register-transfer level is due primarily to the technology -- to MSI and LSI fabrication techniques. However, it is perhaps due also to the creation of systems of primitives at the RT level, first the Macromodules by Wesley Clark (1966) at Washington University and then by one of the authors, (see Bell and Grason, 1971). These have provided the conceptual scheme around which to conceive of design operating exclusively at the register-transfer level, without involvement of the lower logical levels of general combinational and sequential circuits. An additional conceptual tool, in our own case at least, has been the development of an appropriate language for describing RT-level components (so-called PMS notation, see Bell and Newell, 1971). This is a functional notation which has led us off in a somewhat different direction than that implicit in the earlier attempts to create a register-transfer level, e.g., the work of Bartee, which was strongly algebraic in character (essentially formalizing the level as Boolean rings on vector elements).

This book is an attempt to bring this level of design to fruition. We have based it strongly on the use of a specific set of RT-level components, the DEC PDP- 16 modules which we call Register Transfer Modules (RTM's). Our reasons for using RTM's herein are partly pedagogical, partly proprietary (in both a corporate(1) and a personal sense--we like our own modules, they being children


1. The system of modules used here was originally called Register Transfer Modules (RTM's) and was initally designed by one of us (GB) at Carnegie- Mellon University for educational purposes under an NSF equipment grant (GY 5160). They have been described in the literature (Bell and Grason, 1971; Bell, Eggert, Grason, and Williams, 1972). The production version of these modules was developed by John Eggert and Peter Williams of DEC, using the name PDP-16. These differ slightly from the original RTM's, mostly for production reasons. The PDP-16 set is described in the present book, although the precise definition specification is given in DEC PDP-16 Computer Designer's Handbook, 1971.


previous | contents | next