This preview shows page 1. Sign up to view the full content.
Unformatted text preview: to be brought into the cache: 465 ¯ ¯ ¯ The instructions implementing the measurement code (e.g., start counter, get counter, and so on). Let the number of cycles for this be Ñ. The instructions implementing the procedure being measured (procA or procB). Let the number of cycles for this be Ô. The data locations being updated (designated by b1, b2, or b3). Let the number of cycles for this be . Based on the measurements shown in Figure 9.13, give estimates of the values of , Ñ, Ô, and . Given the variations shown in these measurements, a natural question to ask is “Which one is right?” Unfortunately, the answer to this question is not simple. It depends on both the conditions under which our code will actually be used as well as the conditions under which we can get reliable measurements. One problem is that the measurements are not even consistent from one run to the next. The measurement table shown in Figure 9.13 show the data for just one testing run. In repeated tests, we have...
View Full Document
- Spring '10
- The American