Intel Spots A 3888.9% Performance Improvement In The Linux Kernel From One Line Of Code

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • Etherman
    Senior Member
    • Dec 2017
    • 289

    #21
    Originally posted by blackshard View Post

    Tend to agree... news like this are mostly curiosity rather than anything serious
    I find it interesting and news worthy.
    Should I have THP on or off? Maybe this tips the scales?

    Comment

    • fitzie
      Senior Member
      • May 2012
      • 672

      #22
      nice. with this performance boost I can avoid upgrading my machine. even if this isn't representative of my typical load, I'm sure they will find other 1 line speed ups in the future.

      Comment

      • Palu Macil
        Senior Member
        • Feb 2016
        • 123

        #23
        Originally posted by Developer12 View Post
        So intel didn't have any hand in finding the fix or authoring the change, and only found the results the authors expected.

        Why is this news again?
        This article was a bit lower quality than normal…

        Comment

        • erniv2
          Senior Member
          • Apr 2022
          • 266

          #24
          Originally posted by Etherman View Post

          I find it interesting and news worthy.
          Should I have THP on or off? Maybe this tips the scales?
          If you use a common distribution like ubuntu or fedora it is set to madvise (the app has to ask for hugepages) and the only app that comes to mind that uses hugepages in a standard setup i came across is qemu, so no you dont have to even bother thinking about it in your desktop workuse.

          Comment

          • gregzeng
            Senior Member
            • May 2010
            • 170

            #25
            Journalism is journalism. Phoronix used to try having its own type of benchmark tests. There are many benchmark tests for both Linux and Windows now. How do these more stable, more authoritative tests perform?

            The Linux Foundation releases the raw source code every week now. Both revisions of the Linux kernel. Plus new Linux kernels. There are also a few independent compilers of these Linux kernels.

            With so many kernel compiled codes, many specialise in improving Linux performance. Do these actually increase performance, at the expense of other factors?

            There are also many versions of Linux. Does it matter which version? Red Hat, Debian, Linux From Scratch, etc. Similarly there are a few versions of Microsoft Windows. Do these versions also matter? On the many e ways of benching.

            Now that Phoronix is mostly journalism, rather than benching, perhaps my questions could be answered.

            Similarly, do the many variations of the CPU from AMD show big improvements with Linux, instead of Microsoft Windows?

            Comment

            • bug77
              Senior Member
              • Dec 2009
              • 6476

              #26
              Originally posted by moonwalker View Post

              ...the headline IS rather clickbait-y...
              Only if you're new to Phoronix.
              On one hand, there's only so much you can say in a title before it turns into a paragraph. On the other hand, people don't exactly throw money at Michael, so can you really blame him? Me, having been here for a while, I know exactly what to expect when I see a title like that.

              Comment

              • mr_tawan
                Phoronix Member
                • Nov 2014
                • 82

                #27
                Pardon my weak English .... but what exactly does 600% slowdown mean? Like if the original score is 100, then what is the slowdown score?

                Comment

                • ejolson
                  Junior Member
                  • Jun 2015
                  • 3

                  #28
                  Originally posted by V1tol View Post
                  ...now we see if that change makes difference in a real-life workloads
                  The 600 percent regression in real workloads had already been noticed in a numerical differential equations library



                  As mentioned already this library is also used in the cactuBSSN component of the Spec CPU benchmark. The regression has already mysteriously appeared in hardware reviews. For example

                  The 2020s were a fun time for Intel’s Atom line, which went from an afterthought to playing a major role across Intel’s high performance client offerings.


                  Intel’s mobile CPUs have undergone massive changes over the past couple generations as Intel defends its laptop market against AMD, Qualcomm, and to a lesser extent Apple.


                  and there are others.

                  Comment

                  • sfjuocekr
                    Junior Member
                    • Feb 2019
                    • 21

                    #29
                    It is always one line of code...

                    Comment

                    • toves
                      Senior Member
                      • Sep 2021
                      • 120

                      #30
                      Originally posted by mr_tawan View Post
                      Pardon my weak English .... but what exactly does 600% slowdown mean? Like if the original score is 100, then what is the slowdown score?
                      Think of 600% as 6× then 600% slowdown probably means 6 times slower or 1/6 of the original speed.

                      I would have thought this might more normally be stated as 16.67% of the original speed or a 5/6 (83.33%) slowdown.

                      Unfortunately English in the hands of marketing types and Americans is a bit like Humpty Dumpty's words*, it means whatever it suits them to have you think it means.

                      * “When I use a word,’ Humpty Dumpty said in rather a scornful tone, ‘it means just what I choose it to mean — neither more nor less.’
                      'The question is,’ said Alice, ‘whether you can make words mean so many different things.’
                      The question is,’ said Humpty Dumpty, ‘which is to be master — that’s all.”
                      ― Lewis Carroll, Through the Looking Glass

                      Comment

                      Working...
                      X