Announcement

Collapse
No announcement yet.

The ~200 Line Linux Kernel Patch That Does Wonders

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

  • Hey guys, the autogroup patch was updated again by Mike Galbraith to v4 on 2010/11/20.

    You can find the newest patches here: http://git.zen-kernel.org/zen-stable/ - specifically this patch: http://lkml.org/lkml/2010/11/20/91

    Also, for any Debian users - a precompiled performance kernel is maintained at http://liquorix.net/ using a snapshot of the zen-kernel sources with more optimal configuration for desktops/laptops.

    Comment


    • Does anybody have tips how to compile kernel with this patch?
      I tried compiling three different versions of Zen kenel and all failed

      This patch looks like what I'm really looking for Fusion Linux (www.fusionlinux.org)

      Thanks,
      Valent.

      Comment


      • Originally posted by damentz View Post
        Hey guys, the autogroup patch was updated again by Mike Galbraith to v4 on 2010/11/20.

        You can find the newest patches here: http://git.zen-kernel.org/zen-stable/ - specifically this patch: http://lkml.org/lkml/2010/11/20/91

        Also, for any Debian users - a precompiled performance kernel is maintained at http://liquorix.net/ using a snapshot of the zen-kernel sources with more optimal configuration for desktops/laptops.
        WOW !!!!!!!!!!!!!

        This kernel is making video playback better in my Dell Mini 9 running Kubuntu maverick. I can finally run Hulu HD at max resolution, full screen. I am not sure if it's Mike's patch or other tweaks you put in, but this is amazing. Thank you so much. Cheers!

        Comment


        • Originally posted by devius View Post
          How about you finnish/swedish/norwegian-speaking guys start including subtitles in the posts so the rest of us know what you're talking about. Or do I have to start insulting everyone in portuguese?
          http://erasmusv.wordpress.com/2007/0...erse-el-sueco/


          Sätt igång!

          Comment


          • English to Dutch dictionary: 1,5cm thick.
            The same Dutch to English dictionary: 6cm thick.

            My language wins.

            http://translate.google.com/

            Comment


            • Well, currently I've 2 different kernel builds to compare my performance.

              One is an older 2.6.35.9-ZEN kernel build with BFS enabled. The other one, I built yesterday and is a 2.6.36.1 ZEN kernel built with CFS + autogroup patch v4.

              From what I see, there are no major speed improvements in loading times in both kernel versions. (my system with a 4200RPM HD loads Arch in about 27.5s with EXT4, in both kernels).

              At normal loads, it seems BFS has slightly better CPU powersaving than CFS (maybe due to 'Zen Tune' options...?).

              Bu finally, at high loads, I found this patch sligthly better than BFS, specially in cases you're, for instance, tranferring high amounts of data between partitions and watching a full HD video (using CPU rendering). 2.6.35 + BFS at very massive loads stalled a tiny bit, but 2.6.36 + autogroup patches, my system NEVER stalled again...

              But I strongly believe Con Colivas will have a strong answer soon...

              These were my 2c

              Cheers

              Comment


              • Originally posted by RealNC View Post
                -j64 is not heavy load. It's ridiculous load that no scheduler patch should even try to address. Heavy load is something like -j18 on a 6-core machine, and -j6 on a dual-core. -j64 is useless. This is load that your desktop PC will never, ever reach. Not in a billion years.
                Depends on what kind of hardware you're running. Your definition of "a heavy load" is setting your number of GCC threads to 3x the number of cores in the machine. Yes, it would be stupid to run "make -j64" on a netbook or an inexpensive desktop. However, you can get a 24-core dual Opteron workstation for not that much more money than the Core i7 980X desktop system the author used to demonstrate the patch. A "heavy load" on that dual Opteron system is more than 64 threads. I think it is entirely appropriate to test using such a large number of threads since CPUs are multithreaded and getting ever more so as time goes on. It's better to be ahead of the curve rather than behind it, in my opinion.

                Comment


                • Big thanks from the Desktop Users

                  I have not tested the patch but I have used the alternative configuration to the 200 lines patch. What has been told to be quite equivalent...

                  And I can tell that this is one of the biggest progress in the Linux Desktop in the last 2 years. Things are running smoothly even the CPU or I/O is booming and even when the computer is swapping! Simply amazing!!

                  Latest kernels (2.6.32 and 2.6.35 at least) were quite crappy when the free RAM was running low or there was lots of hard disk I/O. This patch/configuration simply changes everything!!

                  Enormous huge thanks from my side!

                  Comment


                  • I'll second that. I'm using gentoo-sources-2.6.32-r2, which has the patch already applied to the sources. It works very well and the machine doesn't have those tiny occasional "hiccups" with the GUI when the disk subsystem is getting hammered. I've never had issues with the system bogging under heavy CPU loads either before or after the patch since this machine has 16 cores and can thus handle pretty high loads regardless of the scheduler.

                    Comment


                    • Originally posted by evolution View Post
                      Bu finally, at high loads, I found this patch sligthly better than BFS, specially in cases you're, for instance, tranferring high amounts of data between partitions and watching a full HD video (using CPU rendering). 2.6.35 + BFS at very massive loads stalled a tiny bit, but 2.6.36 + autogroup patches, my system NEVER stalled again...

                      But I strongly believe Con Colivas will have a strong answer soon...
                      Con has abandoned the whole "automatic task grouping per TTY" idea. On October 6 he released his patch for BFS. 1 month later he said:

                      Those following the development of the patches for interactivity at massive
                      load, I have COMPLETELY DROPPED them as they introduce regressions at normal
                      workloads, and I cannot under any circumstances approve changes to improve
                      behaviour at ridiculous workloads which affect regular ones. I still see
                      precisely zero point at optimising for absurd workloads. Proving how many
                      un-niced jobs you can throw at your kernel compiles is not a measure of one's
                      prowess. It is just a mindless test.
                      Full post: http://ck-hack.blogspot.com/2010/11/...y-comment.html

                      Comment


                      • apparently this patch is being included in opensuse 11.4, due in march 2011 with kernel 2.6.37:

                        https://features.opensuse.org/310920

                        Comment


                        • Originally posted by R3MF View Post
                          apparently this patch is being included in opensuse 11.4, due in march 2011 with kernel 2.6.37:

                          https://features.opensuse.org/310920
                          yes and i believe theyed built it without the bkl... i have m4 installed and 11.4 seems very promising... could be a awesome version on opensuse.

                          Comment


                          • agreed.

                            KDE 4.6
                            kernel 2.6.37
                            kernel patches
                            bretzn app-store
                            blue-devil

                            i'm looking forward to it.

                            Comment


                            • Ok, what's the difference with :
                              nice 10 launch_big_background_job_from_terminal_while_wast ing_time_on_the_web ?!

                              Nevertheless I would reaaaally appreciate some improvements in the system responsiveness at high load under linux : the user should NEVER loose the system control (ie mouse, keyboard responsiveness) whatever the load is. Better stop or renice big processes...

                              On my 5 year old laptop I'm frequently stuck with IO load issues : consider monotasking when updating packages...

                              While watching DVB, if I ever hit 100% ram usage, vlc starts swapping, system freezes and I have to wait 5-10 min for full swap, killed processes and back to normal use :-/

                              Hope this patch is a first step to improve responsiveness under Linux !

                              Comment


                              • Originally posted by blaze33 View Post
                                While watching DVB, if I ever hit 100% ram usage, vlc starts swapping, system freezes and I have to wait 5-10 min for full swap, killed processes and back to normal use :-/

                                Hope this patch is a first step to improve responsiveness under Linux !
                                I hear ya! It's a little weird though. It did not use to be like this in the first half or so of the decade. Did we all move to laptops with crappy IO and/or did the kernel change for the worse in terms of responsiveness?

                                Comment

                                Working...
                                X