Playing around with perf

Posted on August 7, 2012 by Paolo Bernardi

The article was moved to https://www.bernardi.cloud/2012/08/07/playing-around-with-perf/

See you on my new website! 😎

About Paolo Bernardi

Registered software engineer with over 10 years' experience in software system design, cybersecurity and privacy expert, free and open‑source software enthusiast. Certified C2‑level English language user. I live in Acquasparta, an Umbrian town famous for being the seat of the first "Accademia dei Lincei" (Academy of the Lynx‑eyed), together with my beloved wife.
This entry was posted in english, linux and tagged , , , , , , , , , , , . Bookmark the permalink.

5 Responses to Playing around with perf

  1. Kaiwan says:
    December 4, 2012 at 13:57

    Cool! Very informative, thank you!

    I’d like to point out (what i think is) a small typo:
    “…
    At least the Python 3 version seems to fill up the pipeline nicely (1.1 instructions per second), compared to Python 2 (less than 1 instruction per second).
    …”

    I think you meant: “instructions per *cycle* ” in both cases, yes?

    Reply
  2. Gal says:
    May 19, 2013 at 19:06

    Hi! Thanks allot for that!
    Do you happen to know how can I use perf to get a graph of some event as a function of time?
    Say, y-axis is the CPI and x-axis is cycles.

    Any help would be appreciated!

    Reply

  3. Pingback: What are stalled-cycles-frontend and stalled-cycles-backend in 'perf stat' result? | DL-UAT

  4. Jamil says:
    April 10, 2016 at 22:00

    I had a feeling something weird was going on when my Java program was slower than my python version, despite Java being compiled and python being interpreted. Thanks for the informative article.

    Reply

Leave a comment