Announcement

Collapse
No announcement yet.

More Scheduler Optimizations Land In Linux 5.8

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

  • More Scheduler Optimizations Land In Linux 5.8

    Phoronix: More Scheduler Optimizations Land In Linux 5.8

    As part of the many areas of the kernel managed by Ingo Molnar, on Tuesday he submitted the pull request with all of the scheduler code updates for Linux 5.8...

    http://www.phoronix.com/scan.php?pag...-5.8-Scheduler

  • #2
    I really like articles related to process and I/O scheduling, filesystems, inner workings of things etc. Makes me relax especially after grub failed med for the gazillion'th time

    http://www.dirtcellar.net

    Comment


    • #3
      Originally posted by waxhead View Post
      I really like articles related to process and I/O scheduling, filesystems, inner workings of things etc.
      I always read as much of the notes as I can with these sorts of patches. Usually I have no idea what they're talking about. It's still exciting to see them make the system better in a hundred different places and options. Makes me wonder how much improvement there is and in what places (performance, latency, power-efficiency, stability).

      Comment


      • #4
        compared to windows 10 cpu scheduling how good linux is ?

        Comment


        • #5
          Is there a way to tell the Linux kernel scheduler which CPU core has the highest frequency on modern CPUs such as Ryzen 3000?

          Example core frequencies on my CPU:

          Code:
          $ numactl -C 3 perf stat stress-ng --cpu 1
          4.325 GHz
          
          $ numactl -C 12 perf stat stress-ng --cpu 1
          4.409 GHz

          Comment


          • #6
            Originally posted by atomsymbol View Post
            Is there a way to tell the Linux kernel scheduler which CPU core has the highest frequency on modern CPUs such as Ryzen 3000?
            CPPC2 is supposed to handle this, and indeed is used on Windows for this purpose, however is there a Linux scheduler that takes this information into account?

            Comment


            • #7
              Originally posted by numacross View Post

              CPPC2 is supposed to handle this, and indeed is used on Windows for this purpose, however is there a Linux scheduler that takes this information into account?
              CPPC information available in /sys/.../highest_perf appears to accurately reflect the frequencies measured on my Ryzen CPU:

              Code:
              # for((i=0;i<$(nproc);i++)); do echo "CPU $i: $(cat /sys/devices/system/cpu/cpu$i/acpi_cppc/highest_perf)"; done
              CPU 0: 176
              CPU 1: 181
              CPU 2: 171
              CPU 3: 166  # 4.3 GHz max freq
              CPU 4: 196
              CPU 5: 196
              CPU 6: 191
              CPU 7: 186
              CPU 8: 176
              CPU 9: 181
              CPU 10: 171
              CPU 11: 166
              CPU 12: 196  # 4.4 GHz max freq
              CPU 13: 196
              CPU 14: 191
              CPU 15: 186

              Comment


              • #8
                Originally posted by Aryma View Post
                compared to windows 10 cpu scheduling how good linux is ?
                Difficult to answer - it all depends on what you need. Throughput or interactivity. For regular use you would in practice not notice much of a difference unless your system is really heavily loaded in which case Linux really shows it strength.

                But generally speaking Windows process scheduling design (the last time I checked) is in principle more limited, "dumber" and has not changed that much over the years. The improvements is basically (like most Windows things) just bolted on. I suggest you have a look at this:

                https://docs.microsoft.com/en-us/win...ectedfrom=MSDN

                Besides you can have the most efficient scheduler in the world and it would not help you much if the memory management (and in particular swap mechanism) is about as brain dead as the average systemd discussion on phoronix.

                The proper answer to this is that both approaches has advantages and disadvantages. Michael should run some of his benchmarks on some systems with limited RAM and compare both Windows and Linux - I would expect our favorite penguin to take the lead.

                http://www.dirtcellar.net

                Comment

                Working...
                X