Announcement

Collapse
No announcement yet.

Real-Time "PREEMPT_RT" Work Down To Just 50 Patches Atop Linux 6.0-rc1

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

  • Real-Time "PREEMPT_RT" Work Down To Just 50 Patches Atop Linux 6.0-rc1

    Phoronix: Real-Time "PREEMPT_RT" Work Down To Just 50 Patches Atop Linux 6.0-rc1

    Following Sunday's release of Linux 6.0-rc1, yesterday saw the release of 6.0-rc1-rt1 as the set of patches for providing real-time kernel support atop the upstream code-base. There is just roughly 50 patches to go until the PREEMPT_RT functionality is mainlined!..

    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
    Nice! I hope it gets merged soon and more strong efforts take place soon thereafter.

    Comment


    • #3
      Is PREEMPT_RT something that affects input latency as well? I recently tried Fedora Silverblue which had PREEMPT_RT enabled in it's kernel, and I felt that the input latency was lower compared to other distros.

      Comment


      • #4
        Originally posted by user1 View Post
        Is PREEMPT_RT something that affects input latency as well? I recently tried Fedora Silverblue which had PREEMPT_RT enabled in it's kernel, and I felt that the input latency was lower compared to other distros.
        Yes, and no.

        Yes, it can allow much lower latency, but the chief benefit is for things that absolutely must be on time. Like a pacemaker, an airbag trigger or a pulse generatgenerator.
        ​​​​​​The way RT can help here is to run a task that polls a sensor and fill an input queue periodically every 1 ms or so and then let a separate task read the queue every 10 ms.

        Most applications are for very specific domains however.

        Comment


        • #5
          Originally posted by phoronix View Post
          There is just roughly 50 patches to go until the PREEMPT_RT functionality is mainlined!..
          if the number of patches really that significant? You can have 50x one liners or 1x 50 line patch and it would not be much different (especially if the lines are about the same length).

          Not trying to be (too) difficult, but just an attempt of constructive criticism.


          http://www.dirtcellar.net

          Comment


          • #6
            Originally posted by waxhead View Post

            if the number of patches really that significant? You can have 50x one liners or 1x 50 line patch and it would not be much different (especially if the lines are about the same length).

            Not trying to be (too) difficult, but just an attempt of constructive criticism.
            Well, PREEMPT_RT clearly has more than 50 patches before and now it is reduced to 50 patches.
            And the article mentions that many patches simply modifies Kconfig to add new options.

            Comment


            • #7
              There are 4 big patches ~300-600 lines and the rest are less then 100 lines.

              Comment


              • #8
                Originally posted by baxeno View Post
                There are 4 big patches ~300-600 lines and the rest are less then 100 lines.
                Some of those smaller patches can have more weight than the big patches though. As seen with the recent glibc debacle, even a single line can have huge unforeseen ramifications.

                From a quick rundown though, it seems like 7 of those patches simply enable PREEMPT RT for several architectures and another 8 are one-liners (or thereabout). Another five are hacks that will not be required once more robust solutions have been implemented, and five others are Quality of Life improvements. The rest 25 or so patches are the bulk of the rest and indeed take up quite a few lines of code.

                Looks like 6.1 could be it!
                Last edited by wertigon; 18 August 2022, 02:06 PM.

                Comment


                • #9
                  Hey AMD! Please give us a version of a Ryzen with 1 or 2 cores that are fully isolated from the other cores caches for running RT threads.

                  Comment

                  Working...
                  X