This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Some preliminary instrumentation and analyst of systemtap compiling instrumentation scripts
- From: William Cohen <wcohen at redhat dot com>
- To: systemtap at sourceware dot org
- Date: Thu, 4 Oct 2018 12:08:30 -0400
- Subject: Some preliminary instrumentation and analyst of systemtap compiling instrumentation scripts
Hi,
I have been expanding the example how systemtap could be useful. I just added an example that makes use of the systemtap user-space markers to record the amount of time that each phase of systemtap takes:
https://sourceware.org/git/gitweb.cgi?p=systemtap.git;a=blob;f=testsuite/systemtap.examples/apps/stap_time.stp
The goal of this script is to make it easy to collect data on how long it takes for the various example scripts to build and determine if there are any problematic ones that need to be looked at more closely. I have put the raw data, gnuplot scripts, and resulting svg graphs on:
http://people.redhat.com/wcohen/stap_profile/
The readme.txt in the directory describe the files.
One can look at the .svg and then the corresponding .dat to determine which examples have the long elaboration (pass2) and compile (pass4) times.
The reason for the long pass1 times for the x86_64_times.dat is due to the huge number of qemu tapsets on the machine 84MB vs 7.1MB on the guest_x86_64 machine.
I am looking through the graphs and matching up the various peaks with the original *.dat files.
-Will