This preview shows page 1. Sign up to view the full content.
Unformatted text preview: te performance measurements often set up specially conﬁgured machines that minimize any sources of timing irregularity, such as by limiting access and by disabling many OS and networking services. We want methods that application programmers can use on ordinary machines, but there are no widely available tools for this. Instead, we will develop our own. In this presentation we work through the issues systematically. We describe the design and evaluation of a number of experiments that helped us arrive at methods to achieve accurate measurements on a small set of systems. It is unusual to ﬁnd a detailed experimental study in a book at this level. Generally, people expect the ﬁnal answers, not a description of how those answers were determined. In this case, however, we cannot provide deﬁnitive answers on how to measure program execution time for an arbitrary program on an arbitrary system. There are too many variations of timing mechanisms, operating system behaviors, and runtime environment to have one single, simple solution. Instead, we anticipate that you will need to run your own experiments and develop your...
View Full Document
This note was uploaded on 09/02/2010 for the course ELECTRICAL 360 taught by Professor Schultz during the Spring '10 term at BYU.
- Spring '10
- The American