Interval counting is only useful for measuring

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: mechanisms computers use to record the passage of time—one based on a low frequency timer that periodically interrupts the processor and one based on a counter that is incremented every clock cycle. Application programmers can gain access to the first timing mechanism by calling library functions. Cycle timers can be accessed by library functions on some systems, but they require writing assembly code on others. We have deferred the discussion of program timing until now, because it requires understanding aspects of both the CPU hardware and the way the operating system manages process execution. Using the two timing mechanisms, we investigate methods to get reliable measurements of program performance. We will see that timing variations due to context switching tend to be very large and hence must be eliminated. Variations caused by other factors such as cache and branch prediction are generally managed by evaluating program operation under carefully controlled conditions. Generally, we can get accurate measurements for durations...
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