By I. E. Sutherland
Communications of the ACM,
Vol. 32 No. 6, Pages 720-738
The pipeline processor is a common paradigm for very high speed computing machinery. Pipeline processors provide high speed because their separate stages can operate concurrently, much as different people on a manufacturing assembly line work concurrently on material passing down the line. Although the concurrency of pipeline processors makes their design a demanding task, they can be found in graphics processors, in signal processing devices, in integrated circuit components for doing arithmetic, and in the instruction interpretation units and arithmetic operations of general purpose computing machinery.
Because I plan to describe a variety of pipeline processors, I will start by suggesting names for their various forms. Pipeline processors, or more simply just pipelines, operate on data as it passes along them. The latency of a pipeline is a measure of how long it takes a single data value to pass through it. The throughput rate of a pipeline is a measure of how many data values can pass through it per unit time.
Pipelines both store and process data; the storage elements and processing logic in them alternate along their length. I will describe pipelines in their complete form later, but first I will focus on their storage elements alone, stripping away all processing logic. Stripped of all processing logic, any pipeline acts like a series of storage elements through which data can pass.
Pipelines can be clocked or event-driven, depending on whether their parts act in response to some widely-distributed external clock, or act independently whenever local events permit. Some pipelines are inelastic; the amount of data in them is fixed. The input rate and the output rate of an inelastic pipeline must match exactly. Stripped of any processing logic, an inelastic pipeline acts like a shift register. Other pipelines are elastic; the amount of data in them may vary. The input rate and the output rate of an elastic pipeline may differ momentarily because of internal buffering. Stripped of all processing logic, an elastic pipeline becomes a flow-through first-in-first-out memory, or FIFO. FIFOs may be clocked or event-driven; their important property is that they are elastic.
I assign the name micropipeline to a particularly simple form of event-driven elastic pipeline with or without internal processing. The micro part of this name seems appropriate to me because micropipelines contain very simple circuitry, because micropipelines are useful in very short lengths, and because micropipelines are suitable for layout in microelectronic form.
I have chosen micropipelines as the subject of this lecture for three reasons. First, micropipelines are simple and easy to understand. I believe that simple ideas are best, and I find beauty in the simplicity and symmetry of micropipelines. Second, I see confusion surrounding the design of FIFOs. I offer this description of micropipelines in the hope of reducing some of that confusion.
The third reason I have chosen my subject addresses the limitations imposed on us by the clocked-logic conceptual framework now commonly used in the design of digital systems. I believe that this conceptual framework or mind set masks simple and useful structures like micropipelines from our thoughts, structures that are easy to design and apply given a different conceptual framework. Because micropipelines are event-driven, their simplicity is not available within the clocked-logic conceptual framework. I offer this description of micropipelines in the hope of focusing attention on an alternative transition-signalling conceptual framework.
We need a new conceptual framework because the complexity of VLSI technology has now reached the point where design time and design cost often exceed fabrication time and fabrication cost. Moreover, most systems designed today are monolithic and resist mid-life improvement. The transition-signalling conceptual framework offers the opportunity to build up complex systems by hierarchical composition from simpler pieces. The resulting systems are easily modified. I believe that the transition-signalling conceptual framework has much to offer in reducing the design time and cost of complex systems and increasing their useful lifetime. I offer this description of micropipelines as an example of the transition-signalling conceptual framework.
Until recently only a hardy few used the transition-signalling conceptual framework for design because it was too hard. It was nearly impossible to design the small circuits of 10 to 100 transistors that form the elemental building blocks from which complex systems are composed. Moreover, it was difficult to prove anything about the resulting compositions. In the past five years, however, much progress has been made on both fronts. Charles Molnar and his colleagues at Washington University have developed a simple way to design the small basic building blocks . Martin Rem's "VLSI Club" at the Technical University of Eindhoven has been working effectively on the mathematics of event-driven systems [6, 10, 11, 19]. These emerging conceptual tools now make transition signalling a lively candidate for widespread use.
The full text of this article is premium content
No entries found
Log in to Read the Full Article
Please select one of the options below for access to premium content and features.
Create a Web Account
If you are already an ACM member, Communications subscriber, or Digital Library subscriber, please set up a web account to access premium content on this site.
Join the ACM
Become a member to take full advantage of ACM's outstanding computing information resources, networking opportunities, and other benefits.
Subscribe to Communications of the ACM Magazine
Get full access to 50+ years of CACM content and receive the print version of the magazine monthly.
Purchase the Article
Non-members can purchase this article or a copy of the magazine in which it appears.