Announcement

Collapse
No announcement yet.

An Intel Engineer Has Another Optimization For Possible Performance Degradation On Linux

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

  • An Intel Engineer Has Another Optimization For Possible Performance Degradation On Linux

    Phoronix: An Intel Engineer Has Another Optimization For Possible Performance Degradation On Linux

    Besides the long-running FSGSBASE patch series that has the ability to help the performance for CPUs going back years, another engineer on Intel's open-source team has been working on a separate but enticing patch in the name of performance...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    For Possible Performance Degradation
    i don't like where this is going

    Comment


    • #3
      Originally posted by Aryma View Post

      i don't like where this is going
      You don't like preventing possible performance degradation?

      Comment


      • #4
        Originally posted by Volta View Post
        You don't like preventing possible performance degradation?
        Technically, I believe this patch set reverses potentially unnecessary long term performance degradation (huge pages (with their performance benefits) can be reestablished after being broken apart should all the pages again share the same attributes). As is usually the case, not all workloads will benefit, but I can see some long running processes seeing improvements.
        Last edited by CommunityMember; 17 April 2020, 01:50 PM.

        Comment


        • #5
          Originally posted by Volta View Post
          You don't like preventing possible performance degradation?
          Pretty sure he meant it as a joke. After all, intel has a reputation for disregarding security in the name of performance.

          Comment


          • #6
            Originally posted by Volta View Post

            You don't like preventing possible performance degradation?
            Well. Since Linux has zero mandated public performance regression testing...
            Unless proven to fix something and not affecting a lot of other workloads then yes.
            A pretty shitty track record of breaking core performance stuff does exist.

            Adding code to "potentially fix complex problems" can have a lot of side effects.

            Comment


            • #7
              Originally posted by milkylainen View Post

              Well. Since Linux has zero mandated public performance regression testing...
              Unless proven to fix something and not affecting a lot of other workloads then yes.
              A pretty shitty track record of breaking core performance stuff does exist.

              Adding code to "potentially fix complex problems" can have a lot of side effects.
              What OS has such testing? BSD? No. Windows? Not a chance. Windows wasn't even tested before in public and it breaks more than ever with version 10. However, it will be good to have such framework for Linux which has the highest development potential, awesome features and performance.

              Comment

              Working...
              X