Thursday, 16 May 2019

linux-5.1-ck1, MuQSS version 0.192 for linux-5.1

Announcing a new -ck release, 5.1-ck1  with the latest version of the Multiple Queue Skiplist Scheduler, version 0.192. These are patches designed to improve system responsiveness and interactivity with specific emphasis on the desktop, but configurable for any workload.

linux-5.1-ck1:
-ck1 patches:
Git tree:
MuQSS only:
Download:
Git tree:


Web: http://kernel.kolivas.org


This is mostly a resync from 5.0-ck1 with some build and boot fixes courtesy of Serge Belyshev (thanks!)
Enjoy!
お楽しみ下さい
-ck

20 comments:

  1. Thank you very much, sir.

    ReplyDelete
  2. Thank you very much,CK.
    You did a great job!

    ReplyDelete
  3. Smooth as silk, solid as a rock.

    ReplyDelete
  4. Thanks a lot. Running this with Kwin-LowLatency on Gentoo and it feels great.

    ReplyDelete
  5. FYI -
    Each of the tests is specifically designed to trigger a code path in the Linux kernel. Unlike most other benchmarks/performance tests, their purpose is not to simulate a real-life workload. github.com/mfleming/adrestia

    ReplyDelete
  6. Has the AMD/schedutil issue been fixed?

    ReplyDelete
  7. Any chance to get the tunables:
    /proc/sys/kernel/hrtimer_granularity_us
    /proc/sys/kernel/hrtimeout_min_us
    back?

    They were excellent for fine-tuning.

    ReplyDelete
    Replies
    1. They're still there. I'm guessing you just applied muqss? They're part of the -ck patchset.

      Delete
    2. Ah ok.
      My bad.
      Thank you.

      Delete
    3. I switched "back" to -ck now.
      Amazing!

      Delete
  8. Stable on 5.1.15. The new BMQ scheduler shows quicker perf benchmarks but the ck patchset is much more ideal for desktop usage. Thanks!

    ReplyDelete
  9. Anyone else seeing boot time RCU stalls with 5.1.15-ck1? Seems to happen about once every four boots.

    ReplyDelete
    Replies
    1. By the way there's no stack trace for the stalls.

      Delete
    2. In the end this was fixed by adding "nothreadirqs" to the kernel command line.

      P.S. Thanks for the great work CK :)

      Delete
    3. Thanks. That's why I made threaded IRQs not a default configuration since it's not guaranteed to work everywhere.

      Delete
  10. ck1: KSysguard shows 2 cores at 100% when idle, htop shows 2-3% average on all 4 cores but load average of 5.5.
    Is this expected behaviour?

    ReplyDelete
    Replies
    1. Nothing said. Tracked it down to a SSD failure.

      Delete
  11. Please 5.2 me.

    ReplyDelete
  12. When compiling kernel, load goes up to 12 on AMD Ryzen 5 2500U and opening new apps is extremely slow

    ReplyDelete