Announcement

Collapse
No announcement yet.

Two Years With Linux BFS, The Brain Fuck Scheduler

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

  • #21
    Originally posted by RealNC View Post
    Here:

    Me: Kernel 3.0 is unusable for me without BFS. I mean really, a total piece of crap experience. Benchmark THAT.
    And I wanted to know why it's unusable.

    You: Care to explain what's wrong? I know CFS doesn't allow you to use millions Hz as a timer frequency [wtf is that? this is suggesting that CFS doesn't and BFS does], but some people explained it's a dumb stupid idea.
    This is suggesting you can use such high timer with BFS and it's also suggesting such ability is dumb. Con is a maker of BFS and the patch which allows you to use such high frequency and Ingo is a maker of CFS and he didn't make such funny patch.

    Stop trying to weasel out of it with "but I didn't mean that, it's your fault for thinking that". Be a man. Say "I'm sorry, I miscommunicated it, so here's what I meant."
    Don't be kidding me. Didn't I show it's you who had no clue? You were saying I'm FUDing while I just pointed to your problems with KDE. I was talking about your problems with KDE and unstable graphic drivers you were using to point it can be your fault when comes to your CFS problems. You said later BFS fixed them, but you meant it fixed different problems from the future which you mentioned later. Con should make some notice BFS can really brainf*ck your brain...

    So, what is it that you meant? And how does what you meant bear any relevance to this thread (which is about BFS and not some random, highly specialized, special circumstances, HZ altering -ck patch.)
    While it's about BFS it didn't stop you from bashing CFS. Con is a BFS maker and some can consider BFS is some kind of a toy while Con allows you to do such funny things like setting timer to 10KHz. Just a tiny "suggestion".

    Comment


    • #22
      With a name like that, I don't see it grabbing much traction. Just seems like a very ignorant approach.

      Comment


      • #23
        Originally posted by darkbasic View Post
        Jack isn't any better with BFS, you have to enable threaded interrupts and set irq priorities, THAT will make a BIG difference.
        Maybe it's not Jack itself, but the whole audio chain, from the synths up to Jack. Lots of processes involved. With BFS, I can use 64 frames/period, no problem and no audio drops. Total latency of everything combined is well under 6ms. With CFS, it craps out as soon as I actually start playing some synths and CPU load rises. I have to raise latencies up to 15ms to make it work reliably.

        It's really like day and night.
        Last edited by RealNC; 16 August 2011, 09:30 AM.

        Comment


        • #24
          Originally posted by kraftman View Post
          So I understand BFS fixed your slow OpenGL and Kwin crashes? Simply brilliant!*
          If BFS appears to fix kwin crashes, then that could indicate race conditions in kwin.

          Originally posted by RealNC View Post
          Maybe it's not Jack itself, but the whole audio chain, from the synths up to Jack. Lots of processes involved. With BFS, I can use 64 frames/period, no problem and no audio drops. Total latency of everything combined is well under 6ms. With CFS, it craps out as soon as I actually start playing some synths and CPU load rises. I have to raise latencies up to 15ms to make it work reliably.

          It's really like day and night.
          Did you try Linux 3.0 with automatic cgroups enabled? It is very good in that release. It should improve that situation.

          Comment


          • #25
            I don't get why so many people are talking about these benchmarks being "useless" because they are missing the point of BFS.

            I think, it is interesting to see how much raw performance you sacrifice (or even gain) when changing to BFS. These benchmarks suggest it's mostly a pretty good deal. Just maybe not for apache.

            Comment


            • #26
              I stopped using BFS when i did benchmarks with i7-880. With an E8400 BFS improved compile speed, but with HT it did not matter at all if BFS was used or not. Also desktop performance improved a lot with mainline kernels. Basically i see no specific reason to patch a kernel, maybe increase HZ for better reaction time, but thats all.

              Comment


              • #27
                Originally posted by Shining Arcanine View Post
                Did you try Linux 3.0 with automatic cgroups enabled? It is very good in that release. It should improve that situation.
                Yep, automatic grouping was enabled.

                Comment


                • #28
                  Originally posted by Kano View Post
                  I stopped using BFS when i did benchmarks with i7-880. With an E8400 BFS improved compile speed, but with HT it did not matter at all if BFS was used or not. Also desktop performance improved a lot with mainline kernels. Basically i see no specific reason to patch a kernel, maybe increase HZ for better reaction time, but thats all.
                  Yeah, but that's just people using powerful systems. Try a 2.4GHz E6600. Without BFS the GUI gets really bad.

                  Comment


                  • #29
                    Originally posted by yoshi314 View Post
                    it's just too bad that BFS + systemd is still a no go :/ (no cgroup support).
                    Not sure what do you mean, i have been using BFS + systemd since almost a year ago without problems.

                    And yes, i have cgroups enabled (what bfs doesn't support is cpu load balancing inside the cgroup and other things, who isn't needed if you are using BFS anyway).

                    Comment


                    • #30
                      I've been using BFS for about a year on one of my machines, and yes it certainly feels more responsive when I do many cpu intensive tasks (rendering, encoding etc), however I'm using two other systems aswell with cfs and it's not like they are in any way unuseable, even under heavy load. So can we please pull back a little on the exaggerations? Also, as for the kernel shipping with cfs, is that surprising? AFAIK the main areas where Linux is used is where throughput is favoured over responsiveness. Having both schedulers available is great, so why fight!?

                      Comment

                      Working...
                      X