In chapter 6 Section 2 two sets of perf data are compared. The task is to analyze both data files to determine differences that would indicate which of the programs is running most efficiently.
Then after showing the commands for perf report, Step 3.6 says "After comparing the event listings, you will discover that program1 is the more efficient running progtram of the two".
The instructor in the video just parrots this statement without any other details.
I am not sure on what basis (what exact numbers are being compared and how) this conclusion is made.
Thanks in advance for any details you can throw on this topic.
Subu
I agree that more context can be added to the GE so as to explain the criteria or the basis with which it was decided that program1 is more efficient than program2 -
However, when you look at step 3 : it indicates that you need to compare the two programs for each of the event listings to arrive at that conclusion :
So when you proceed as per step 3.5 by scrolling up and down ( in side by side terminal windows to compare the events overhead ) - you will see the following and then you can quickly decide which of the program looks more efficient ( by comparing important event overheads ) :
branch-misses fr program2 :
branch- loads and branch-loads-misses - program1
branch- loads and branch-loads-misses - program2 :
cache misses and instruction efficiency - program 1 & program 2
https://perf.wiki.kernel.org/index.php/Tutorial#Sample_analysis_with_perf_report
I will raise a feedback ( IIRC I had already raised one for this ) JIRA on this to the appro[riate team to add a more detail onto this.
Thank you for reporting this to us.
Red Hat
Learning Community
A collaborative learning environment, enabling open source skill development.