Messages in this thread | | | Date | Tue, 20 Mar 2012 15:04:10 -0500 | Subject | Re: [PATCH 1/1] Fix number of events displayed in header | From | Ashay Rane <> |
| |
Oh, okay. Thanks for the clarification!
However, my reasoning behind including the actual count of the events was that (as far as I can tell) it is not displayed in any of the perf report outputs. I understand that the count is not precise (because of sampling errors) but the sampling frequency (-F) can always be adjusted to get a more accurate number.
With the inclusion of the count, it then becomes possible to compare multiple different versions of a program. Hence the patch.
Ashay
2012/3/20 Peter Zijlstra <a.p.zijlstra@chello.nl>: > On Tue, 2012-03-20 at 14:51 -0500, Ashay Rane wrote: >> The problem I am seeing is that the `perf record' output does not >> comform with the output of `perf stat'. For example, for `hackbench 10 >> process 1000', I see 19 x 10^9 cycles reported by `perf stat'. >> However, `perf report -n' prints "Events: 2K cycles" and `perf report >> -n --dso hackbench' prints "Events: 47 cycles". > > Ah, read it as if perf report is saying: Samples: 47 of event: cycles. > > That is, its telling you the cycles event had 47 overflows causing 47 > samples to be taken (IP, etc..) and written out. > > The point of reporting this number is that it puts the accuracy of the > percentages into perspective. The more samples, the greater the coverage > and more accurate the actual percentage of time spend in the various > functions.
| |