Announcement

Collapse
No announcement yet.

An Early Look At Linux 4.16 Performance On Five Systems

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

  • An Early Look At Linux 4.16 Performance On Five Systems

    Phoronix: An Early Look At Linux 4.16 Performance On Five Systems

    Here are some preliminary benchmarks of the Linux 4.16 development kernel compared to Linux 4.15 stable on five different systems.

    http://www.phoronix.com/vr.php?view=26027

  • #2
    Typo?:

    Originally posted by phoronix View Post
    This testing is just intended as a disverse range of tests to compare Linux 4.15 stable

    Comment


    • #3
      those charts' legend text for yellow is very hard to read !

      Comment


      • #4
        Originally posted by eva2000 View Post
        those charts' legend text for yellow is very hard to read !
        _+1_ …

        Comment


        • #5
          Page 1: I wonder if those regressions are present if you turn off all Spectre/Meltdown mitigations. They seem to be appearing near when whatever iteration of retpoline was pushed for the masses. Ubuntu is especially late on these patches. As these mitigations are still being worked on, "pure" kernel performance is hard to be seen. Performance may go up and down any moment.

          Comment


          • #6
            Originally posted by eydee View Post
            Page 1: I wonder if those regressions are present if you turn off all Spectre/Meltdown mitigations. They seem to be appearing near when whatever iteration of retpoline was pushed for the masses. Ubuntu is especially late on these patches. As these mitigations are still being worked on, "pure" kernel performance is hard to be seen. Performance may go up and down any moment.
            Initial Retpoline support for x86_64 landed last month, not at the start of February, albeit other Retpo/kpti improvements in 4.16.
            Michael Larabel
            http://www.michaellarabel.com/

            Comment


            • #7
              No...the yelllow text is NOT hard to read. It's IMPOSSIBLE to read. Might want to rethink that color Michael. Thanks for the testing none the less. Very interesting.

              Comment


              • #8
                Originally posted by Jumbotron View Post
                No...the yelllow text is NOT hard to read. It's IMPOSSIBLE to read. Might want to rethink that color Michael. Thanks for the testing none the less. Very interesting.
                I didn't come up with the color. It was auto-generated by PTS when running out of other colors, albeit should enhance that logic to try to come up with a better color selection algorithm but it's a constant struggle.
                Michael Larabel
                http://www.michaellarabel.com/

                Comment


                • #9
                  Michael, I can read it, but if I were to change it I'd use matplotlib's new color tableau:
                  https://github.com/matplotlib/matplo...or_data.py#L19
                  and then cycle over linestyles/hatch-pattern when running out of colors.

                  EDIT: impressive work as always!

                  Comment


                  • #10
                    How did the i5 beat the i7 in many tests?

                    Comment

                    Working...
                    X