Announcement

Collapse
No announcement yet.

The v2 Rotary Interactivity Favor Scheduler

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

  • #11
    3.3.5 with RIFS

    Comment


    • #12
      Originally posted by Naib View Post
      The lagging problem was gone after reboot and switching back to my backup kernel
      If you have time please try to boot with 3.3.x-RIFS and It is more easy for.me to figure out what prpblems RIFS has.

      Comment


      • #13
        Originally posted by Naib View Post
        3.3.5 with RIFS
        Thanks. Does it handle imteractivity better than BFS with high workload ?For example is Firefox smooth even under high workload eg.make -j128 or even crazier like make -j512?
        Last edited by 3766691; 15 May 2012, 10:07 AM.

        Comment


        • #14
          Originally posted by Naib View Post
          3.3.5 with RIFS
          Thanks. Does it handle imteractivity better than BFS with high workload ?For example is Firefox smooth even under high workload eg.make -j128 or even crazier like make -j512?

          my sucky Android duplicate my message
          Last edited by 3766691; 15 May 2012, 10:19 AM.

          Comment


          • #15
            Where can we ask for support? I can't compile the kernel with this patch, must be doing something wrong (kind of new to the whole custom kernel scene)

            Comment


            • #16
              Originally posted by elmariachi View Post
              Where can we ask for support? I can't compile the kernel with this patch, must be doing something wrong (kind of new to the whole custom kernel scene)
              Oops.Could you send me your config.Also this patch is just availale for 3.3.x kernel right now

              Comment


              • #17
                It was me I had conflicting patches. Applied to vanilla 3.3.5 worked. It seems snappier, but that can also be placebo, I'll have to test some more. I also notice the lag when switching between apps, but I guess being snappy everywhere all the time is kind of utopic, no?

                Time to compile pacworld

                Comment


                • #18
                  Originally posted by elmariachi View Post
                  It was me I had conflicting patches. Applied to vanilla 3.3.5 worked. It seems snappier, but that can also be placebo, I'll have to test some more. I also notice the lag when switching between apps, but I guess being snappy everywhere all the time is kind of utopic, no?

                  Time to compile pacworld

                  Thanks for giving comment. The reason why it performs snappier is because it don't use complex and unaccurate algorithum to give priority to interactive task.
                  I 'm going to try using the algorithum RIFS has used and EEVDF to priovide a more fairer scheduling. The switches between user-interactive task is smoother than BFS with high workload. It seems lag when switching to a cpu intensive task from interactive task. ;-)
                  Last edited by 3766691; 16 May 2012, 10:00 AM.

                  Comment


                  • #19
                    Are you the main author of RIFS?
                    forgot to say: I'm using it with deadline, it seems to be faster than BFQ or CFQ. Noop and deadline are faster for me (normal hdd, non-sdd)

                    funny thing: governor set to performance, no dyn ticks, nothing to save energy on, and I get almost exactly the same temperature as with BFS with everything on "eco" mode... maybe it's not the scheduler, but it's funny
                    Last edited by elmariachi; 16 May 2012, 03:11 PM.

                    Comment


                    • #20
                      Originally posted by elmariachi View Post
                      Are you the main author of RIFS?
                      forgot to say: I'm using it with deadline, it seems to be faster than BFQ or CFQ. Noop and deadline are faster for me (normal hdd, non-sdd)
                      Yes,

                      Originally posted by elmariachi View Post
                      funny thing: governor set to performance, no dyn ticks, nothing to save energy on, and I get almost exactly the same temperature as with BFS with everything on "eco" mode... maybe it's not the scheduler, but it's funny
                      There are not much relationship between a CPU scheduler and the temperature.
                      Also, RIFS handles interactivity better than BFS with high workload. You can just try make -j128 and play your music or browse webpage with Firefox. According to my test result, 3.3.1 with BFS hanged my box with make -j512, and the music stopped for more than 1 minute, but with BFS it won't.

                      Comment

                      Working...
                      X