This preview shows page 1. Sign up to view the full content.
Unformatted text preview: ve this value to the appropriate position in the array. This process continues until either the error criterion is satisﬁed, in which case we indicate that the measurements have “converged,” or we exceed the limit Å , in which case we indicate that the measurements failed to converge. Experimental Evaluation We conducted a series of experiments to test the accuracy of the Ã -best measurement scheme. Some issues we wished to determine were: 1. Does this scheme produce accurate measurements? 2. When and how quickly do the measurements converge? 3. Can the scheme determine the accuracy of its own measurements? One challenge in designing such an experiment is to know the actual run times of the programs we are trying to measure. Only then can we determine the accuracy of our measurements. We know that our cycle timer gives accurate results as long as the computation we are measuring do not get interrupted. The likelihood of an interrupt is small for computations that are much shorter than the timer interval and...
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