Announcement

Collapse
No announcement yet.

Linux 5.5 Seeing Some Wild Swings In Performance - Improvements But Also Regressions

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by Michael View Post

    Right some tests are ~50% faster, others ~50% slower. I am working on bisecting the biggest ones right now.
    Erm, what about making kernel 5.4 performance a baseline (100%) and showing 5.5 gains/losses vs. the old kernel? That would be a lot less confusing.

    Comment


    • #12
      perpetually high aufkrawall

      acpi-cpufreq really seems to have two advantages over intel_pstate & intel-cpufreq:

      1.
      It doesn't rely on the Intel idle driver while still being able to put the CPU into its deepest sleep state. (C6 in my case -
      i5-3350P [IvyBridge])
      This You can easily check with:
      Code:
      sudo cpupower monitor
      2.
      It isn't misleading as opposed to Intel's performance governor:
      acpi-cpufreq's performance setting always requests the highest frequency from the CPU (usually the boost state), whereas Intel's very own driver still dynamically scales the frequency, which also leads to increased latency.

      Comment


      • #13
        Wow. Epic swings. That is just rubbish. Performance regression on directory basis?
        How hard can it be? Run on a farm. Compile on every commit, build against a few targets, vote for a suite, run it and dump a feedback on the developer of the commit. Priority to specific directories in the kernel. arch/*, block, kernel, mm, fs.
        Last edited by milkylainen; 12-01-2019, 06:29 PM.

        Comment


        • #14
          Originally posted by Michael View Post

          Right some tests are ~50% faster, others ~50% slower. I am working on bisecting the biggest ones right now.
          I find it interesting(was not pointed out) that on Intel "Memcached mcperf" sees a big regression while on AMD it sees a big performance boost.

          Comment


          • #15
            Originally posted by birdie View Post

            Erm, what about making kernel 5.4 performance a baseline (100%) and showing 5.5 gains/losses vs. the old kernel? That would be a lot less confusing.
            Patches always welcome for improving result presentation... I try to work on new visualization improvements as time allows.
            Michael Larabel
            http://www.michaellarabel.com/

            Comment


            • #16
              Originally posted by Alliancemd View Post

              I find it interesting(was not pointed out) that on Intel "Memcached mcperf" sees a big regression while on AMD it sees a big performance boost.
              This was just a one-page summary quick synopsis... The future multi-page articles will have more detail.
              Michael Larabel
              http://www.michaellarabel.com/

              Comment


              • #17
                Bad news for linus

                Comment


                • #18
                  Originally posted by Michael View Post

                  Right some tests are ~50% faster, others ~50% slower. I am working on bisecting the biggest ones right now.
                  there was something in the mailing list about a performance bisect honeypot in the last week

                  Comment


                  • #19
                    Originally posted by milkylainen View Post
                    Wow. Epic swings. That is just rubbish. Performance regression on directory basis?
                    How hard can it be? Run on a farm. Compile on every commit, build against a few targets, vote for a suite, run it and dump a feedback on the developer of the commit. Priority to specific directories in the kernel. arch/*, block, kernel, mm, fs.
                    Looks like Intel runs something like that, see https://lkml.org/lkml/2019/11/26/748

                    Actually this robot reports a lot of various regressions along with git commits which caused them.

                    Kudos to Intel for doing that. The company really does great things for Linux (and as far as I know they extensively use Linux for developing new CPU architectures).

                    Comment


                    • #20
                      Originally posted by Linuxxx View Post

                      I know how You feel...

                      However, we as Intel victims are not completely helpless:

                      Recently, I have found out why passing
                      Code:
                      intel_pstate=disable
                      to the kernel leads to such drastic improvements in performance & latency:

                      From the RedHat Low-Latency tuning guide:


                      So all along it was Intel's IDLE driver that was the culprit, not the P-State driver!
                      This finally explains why Intel systems always felt so sluggish, even though I was always using the Performance governor while also setting the so-called Perf-Bias to zero for maximum performance-bias!

                      Seriously Intel, are you doing this on purpose?
                      I haven't used the P-State frequency driver in years because I feel it's just a big lie.
                      powersave does nothing to save power, and performance does nothing to help with performance!

                      Comment

                      Working...
                      X