Chapter05Rev08 - Chapter 5 (Revision number 8) Processor...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
1 Chapter 5 (Revision number 8) Processor Performance We have seen the basics of instruction-set design and the implementation of the processor. Processor design and implementation is a quantitative exercise. An architect is constantly evaluating his/her decision to include an architectural feature in terms of its impact on performance. In this chapter, we will explore the performance metrics that are of relevance in processor design. 5.1 Space and Time Metrics Due to the hype about processor speed, we always think of “MHz” “GHz” and “THz” whenever we think of processor performance. These terms, of course, refer to the clock cycle time of the processor. We know from Chapter 3 that the clock cycle time is determined by estimating the worst-case delay in carrying out the datapath actions in a single clock cycle. Is processor speed the only determinant of the processor performance? Let us say you wrote a program “foo” and are running it on a processor. What performance metrics would be of interest to you? Clearly, the two things that would interest you are, how much memory is foo occupying ( space metric ) and how long does it take to run foo ( time metric ). Memory footprint quantifies the space metric, while execution time quantifies the time metric. We define the former as the space occupied by a given program and the latter as the running time of the program. Do these metrics have anything to do with what we have seen about processor design so far? The instruction-set architecture of the processor has a bearing on the memory footprint of the program. First, let us understand the relationship between the two metrics. There is a belief that the smaller the footprint the better the execution time. Do you agree? The conventional wisdom in the 70’s was that this premise is true and led to the design of Complex Instruction Set Computer (CISC) architectures. However, the advent of Reduced Instruction Set Computer Architecture in the 80’s debunked this myth to a great extent. We will not get into the CISC vs. RISC debate! However, there is not a strong correlation between the memory footprint and the execution time. Just as a pathological example, consider a program that consists of a million instructions. Suppose there is a tight loop in the program with just 10 instructions where most of the action is, then clearly the size of the program does not matter! Let us try to understand what determines the execution time of the program. The number of instructions executed by the processor for this program is one component of the execution time. The second component is the number of clock cycles taken by each of these instructions (usually referred to as CPI clocks per instruction ). The third component is the clock cycle time of the processor.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
2 So, Execution time = Number of instructions executed * CPI Avg * clock cycle time Of course, if we want to be exact we can substitute the first two terms in the above equation by the summation of the product of the number of instructions of a particular
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 11/25/2010 for the course CENG 100 taught by Professor Ceng during the Spring '10 term at Universidad Europea de Madrid.

Page1 / 11

Chapter05Rev08 - Chapter 5 (Revision number 8) Processor...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online