This is the html version of the file
automatically generates html versions of documents as we crawl the web.
Chapter 9 Instructor's
Chapter 9, Alternative Architectures, provides an overview of alternative architectures that
have emerged in recent years. RISC, Flynn's Taxonomy, parallel processors, instruction-level
parallelism, multiprocessors, interconnection networks, shared memory systems, cache
coherence, memory models, superscalar machines, neural networks, systolic architectures,
dataflow computers, and distributed architectures are covered. Our main objective in this
chapter is to help the reader realize we are not limited to the von Neumann architecture, and to
force the reader to consider performance issues, setting the stage for the next chapter.
Lectures should focus on the following points:
RISC architectures were introduced in Chapters 4 and 5. This chapter
goes into more detail, covering the RISC versus CISC debate, speedup, and overlapping
This is the most widely accepted method for classifying computer
architectures. It is based on the number of instructions and the number of data streams,
and includes SISD, SIMD, MISD, and MIMD.
Parallel and multiprocessor architectures.
This section covers SIMD and MIMD
architectures, including superscalar, VLIW, and vector processors, with examples. Both
shared memory machines (SMM, UMA, and NUMA) and interconnection networks are
It is important to understand how the processors in MIMD
systems communicate with the various memories and with each other. Interconnection
networks are often used, and are categorized according to topology. Completely connected,
star, linear, ring, mesh, tree, and hypercube are covered, and examples of each are given.
In addition, dynamic networks (using 2x2 switches and stages) are also introduced.
Alternative parallel processing approaches.
Although the traditional von Neumann
architecture is a popular platform for study, it is important to cover alternatives to this
architecture to gain a better overall understanding of computer architecture. Indeed, some
highly complex problems cannot be solved using our traditional model of computation .
The Essentials of Computer Organization and Architecture