The rst eld encodes the total number of seconds that

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: hes in this code, and hence we conclude that the variations must be due to cache effects. 464 CHAPTER 9. MEASURING PROGRAM EXECUTION TIME Measurement Examples: Small Array 10 9 8 7 Time (ms) 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Sample Load 1 Load 4 Figure 9.12: Measurements of Short Duration Procedure under Different Loading Conditions The variations are not as extreme as they were in Figure 9.11, but they are still unacceptably large. Measurement 1 2 3 4 5 6 Call procA(b1) procA(b2) procA(b3) procA(b1) procB(b1) procB(b2) Cycles 399 132 134 100 317 100 Figure 9.13: Measurement Sequence with Identical Procedures Operating on Identical Data Sets. The variations in these measurements are due to different miss conditions in the instruction and data caches. 9.4. MEASURING PROGRAM EXECUTION TIME WITH CYCLE COUNTERS Practice Problem 9.6: Let be the number of cycles that would be required by a call to procA or procB if there were no cache misses. For each computation, the cycles wasted due to cache misses can be apportioned between the different items needing...
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