The inherent inaccuracies of the interval counting

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: mp; [2] 5037 Running foo 200 & > bg 5035 5035: No such process > bg 5036 [1] 5036 foo 100 & > /bin/kill 5036 Job 5036 terminated by signal: Terminated Wait for fg job to finish. > fg %2 > quit unix> Back to the Unix shell Figure 8.36: Sample shell session for Problem 8.20. Chapter 9 Measuring Program Execution Time One common question people ask is “How fast does Program run on Machine ?” Such a question might be raised by a programmer trying to optimize program performance, or by a customer trying to decide which machine to buy. In our earlier discussion of performance optimization (Chapter 5), we assumed this question could be answered with perfect accuracy. We were trying to establish the cycles per element (CPE) measure for programs down to two decimal places. This requires an accuracy of 0.1% for a procedure having a CPE of 10. In this chapter, we address this problem and discover that it is surprisingly complex. You might expect that making near-perfect timing measurements on a computer system would be straightforward. After all,...
View Full Document

Ask a homework question - tutors are online