|
|
|
|
|
|
|
Chapter 10
Processor Studies |
|
|
|
|
|
|
|
|
Good design is the product of much work. The effort involves a team of market analysts, competitive product analysts, designers, technologists, and software engineers, directed by management and financial considerations. Design is a constant series of tradeoffs between functional specifications and market requirements vs. functionality, cost, and performance. The studies we present in this chapter clearly cannot capture the man-years of effort that typify the best in processor design. The design situations we present attempt to introduce the reader to simplified design situations without making the problems overly contrived. In this chapter, we present two extended design studies. We begin with the functional specifications. These are the engineering assumptions, market considerations, and determinations that are presented to the designer as the design is begun. As in the rest of this text, we focus on the design issues in instruction sequencing in the memory hierarchy. We allocate a fraction of the design to execution resources and expect performance proportional to the amount of processor resources so dedicated. |
|
|
|
|
|
|
|
|
10.1 The Baseline Mark II |
|
|
|
|
|
|
|
|
This study details the design of a load/store processor to be used as the core of a high-performance workstation, the Baseline Mark II. This workstation is an improved version of the original Baseline Mark I, the only significant change being the replacement of the processor. Some minor changes to the system are planned, but none that are fundamental to the processor or system architecture, in order to maximize compatibility between the systems and to allow an easy upward migration path for customers who already own the original Baseline Mark I. |
|
|
|
|
|
|
|
|
The Baseline-series workstations are targeted at the scientific and engineering communities, and are expected to be used in environments where a significant amount of the workload will be simulation and analysisonly a nominal portion of the workload will be "general-purpose applications" such as editors, databases, and so forth. Users of these machines are interested in maintaining their installed base of applications and having them run faster with newer machines. The Mark II unit is seen as meeting their needs. This study, therefore, is an extension to studies 2.3 and 4.11. |
|
|
|
|
|