Announcement

Collapse
No announcement yet.

Con Kolivas Announces First Major Release Of MuQSS, Successor To BFS

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

  • Con Kolivas Announces First Major Release Of MuQSS, Successor To BFS

    Phoronix: Con Kolivas Announces First Major Release Of MuQSS, Successor To BFS

    At the beginning of the month well-known independent kernel contributor Con Kolivas confirmed he was working on a new project called MuQSS as an evolutionary successor to his Brain F*** Scheduler. This Saturday morning he's now announced the first stable major release of MuQSS...

    http://www.phoronix.com/scan.php?pag...-Major-Release

  • #2
    massive rewrite of BFS designed to maintain the same interactivity and responsiveness using the same algorithm for scheduling decisions, along with a very simple overall design that is easy to understand, model, and hack on
    I don't lnow, if it's really as simple as described, where does the massive rewrite come from?

    Comment


    • #3
      I read this as: "Has the same algorthm etc. as BFS but additionally having a simple design".

      Comment


      • #4
        Lost all interest to scheduler.
        It has 0 impact on gaming.
        Did feckloads of Unigine Valley performance tests with various schedulers, including BFS - 0 difference.
        Tested on FX8350 and FX6350 AMD processors, 6 and 8 core CPU's with max speed on one core of 4.4GHZ.

        Comment


        • #5
          Originally posted by dimko View Post
          Lost all interest to scheduler.
          It has 0 impact on gaming.
          Did feckloads of Unigine Valley performance tests with various schedulers, including BFS - 0 difference.
          Tested on FX8350 and FX6350 AMD processors, 6 and 8 core CPU's with max speed on one core of 4.4GHZ.
          that's a foolish assumption. Minimum frame rates improved in previous benchmarks I've seen and the ck patch was the only way i could keep a responsive desktop during heavy makepkg operations prior to the xorg multi threading patches. this also implies that during heavy load the ck patch will improve input responsiveness directly impacting gaming.

          Comment


          • #6
            Originally posted by dimko View Post
            It has 0 impact on gaming.
            So you get the same score/fps by say compiling the kernel in the background while running Valley, on both CFS and Con's BFS, right?

            Comment


            • #7
              Originally posted by dimko View Post
              Lost all interest to scheduler.
              It has 0 impact on gaming.
              Did feckloads of Unigine Valley performance tests with various schedulers, including BFS - 0 difference.
              Tested on FX8350 and FX6350 AMD processors, 6 and 8 core CPU's with max speed on one core of 4.4GHZ.
              Did it make any difference to minimum fps? I expect this scheduler to have worse max peak fps and max average fps but better minimum fps, which I imagine would be priority for tournament machines.

              Comment


              • #8
                Originally posted by dimko View Post
                Lost all interest to scheduler.
                It has 0 impact on gaming.
                Did feckloads of Unigine Valley performance tests with various schedulers, including BFS - 0 difference.
                Tested on FX8350 and FX6350 AMD processors, 6 and 8 core CPU's with max speed on one core of 4.4GHZ.
                The scheduler is about multitasking (managing multiple processes running on your computer), not about the performance of a single process.

                If you just play a game and the game is the only thing actively wanting to use your CPU, then what scheduler you use should make no difference, exactly as you said (if it does, then something is *very* wrong).

                However, if you are running multiple things at once (such as background tasks (compiling, some other calculation, etc) using your CPU + the foreground game or app you are using), then the scheduler will make a difference. If you want to be able to have background tasks running, loading up your CPU, and still be able to play games / listen to music / browse the web / whatever, while all of that is happening, then BFS *will* make a difference. BFS brings you better responsiveness and interactivity (so you feel like your computer is running smoother and responds faster to input), at the cost of lower actual performance/throughput for the processes, compared to the stock kernel.

                So, if you try to compile the kernel or run some CPU benchmark or other heavy load in the background, and play a game at the same time:

                - Stock kernel: your background compilation will complete slightly faster, and your background benchmark will get a slightly higher score, and your game might get slightly higher *average* FPS, but it will be stuttery and inconsistent, and your input might lag, and you will overall not have a very good experience.

                - BFS: your background compilation will complete slightly slower (but very slightly), and your background benchmark will get a slightly lower score, and maybe your game will get slightly lower *average* FPS, but it will be more consistent, not as stuttery, and your input will feel more responsive, making for a better overall experience.

                Comment


                • #9
                  Originally posted by Licaon View Post
                  So you get the same score/fps by say compiling the kernel in the background while running Valley, on both CFS and Con's BFS, right?
                  nobody cares about gaming performance while compiling
                  this is another 4096 core circlejerk while 70% of desktop users(nvidia) don't even have working drivers out of the box on linux

                  Comment


                  • #10
                    Since CFS got CONFIG_SCHED_AUTOGROUP the difference between BFS and CFS isn't that big anymore

                    Comment

                    Working...
                    X