Observe that in all three cases the measurements for

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: s of experiments that compared the time ÌÑ measured by the operating system for a sample computation versus our estimate of what the time Ì would be if the system resources were dedicated solely to performing this computation. In general, Ì will differ from ÌÑ for several reasons: 1. The inherent inaccuracies of the interval counting scheme can cause than Ì . Ì Ñ to be either less or greater 2. The kernel activity caused by the timer interrupt consumes 4 to 5% of the total CPU cycles, but these cycles are not accounted for properly. As can be seen in the trace illustrated in Figure 9.4, this activity finishes before the next timer interrupt and hence does not get counted explicitly. Instead, it simply 9.3. CYCLE COUNTERS 459 reduces the number of cycles available for the process executing during the next time interval. This will tend to increase ÌÑ relative to Ì . 3. When the processor switches from one task to another, the cache tends to perform poorly for a transient period until t...
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.

Ask a homework question - tutors are online