Announcement

Collapse
No announcement yet.

Torvalds' Comments On Linux Scheduler Woes: "Pure Garbage"

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

  • #81
    In the original author's blog post (link), he claims:
    I spent months trying to find good benchmarks, (still not entirely successful) writing my own spinlocks, mutexes and condition variables and even contributing a patch to the Linux kernel.
    Does anyone know what the patch was for? Was it accepted?

    Also, it's kinda sad it took him months to realize that userspace spinlocks were a bad idea. And even then, I think it was only the data that convinced him.

    Comment


    • #82
      Originally posted by vb_linux View Post
      In different threads I have seen this "Google Developer" thing. From what I understood, he is game developer working for a studio targeting Google Stadia. Huge difference.
      Yeah, I was wondering about that. I didn't see anything in his post or comments to indicate that he worked for Google.

      Comment


      • #83
        Originally posted by coder View Post
        In the original author's blog post (link), he claims:

        Does anyone know what the patch was for? Was it accepted?
        https://lkml.org/lkml/2019/12/4/1373

        Comment


        • #84
          Originally posted by PreparationH67 View Post

          Maybe your communications skills just suck and you have too much of an ego to improve in that manner.
          Or maybe you're assuming I was talking about me?

          Comment


          • #85
            Originally posted by brainlet_pederson
            Given all the bloviating and humble bragging in his blog post, it was probably just a lie. Bet he wasn't counting on Linus calling his bluff.
            It'd be pretty dumb to lie about something so specific and easily verifiable. Whatever you think of that guy, I wouldn't demean him by putting him on the same level as the president of the USA.

            Anyway, PuckPoltergeist seems to have found it (see post #84).

            P.S. I also wouldn't say he was bluffing. He made some incorrect conclusions, but he seemed to believe them. Bluffing is when you're putting forth a view that you know to be false.
            Last edited by coder; 01-07-2020, 11:00 PM.

            Comment


            • #86
              As Linus Torvalds himself already said: "Reality is messy!"

              Same applies to using a mutex for locking.

              Here is what Philip Rebohle [a.k.a. cute2dgirl / before you ask: he obviously likes anime] (the man behind DXVK) had to say on the matter:
              I am obviously aware of that whole situation, and nobody really is a big fan of spinlocks since they are known to be shit when contested, but we have some locks which can be locked several thousand times per frame, and the overhead of locking an std::mutex (especially on wine) is so high that you can easily lose 15-20% performance in the process, for no practical benefit.
              https://github.com/doitsujin/dxvk/co...80b4a#comments

              Also Michael
              I believe it would be interesting to do another round of CPU governor benchmarks, as it seems the major changes for Linux 5.5 were especially made for appealing to schedutil (see for example the heavier reliance on PELT [Per-Entity Load Tracking]).

              Comment


              • #87
                People say that such comments contribute to a "toxic" development environment. However, they can't criticize it without turning this into a crazy SJW witch hunt and demanding heads to be rolled. As such, they are not constructive at all and make any environment even more toxic than the people they criticize.

                However, I think that using such schedulers on Linux would be the easier path to port a game, as Windows schedulers seem to be widely used in game development. As such, it would no long be necessary to translate entire code segments into something completely different just to get the same performance of the original Windows version. It would save countless work hours and ultimately increase the quantity of software available on Linux.

                Comment


                • #88
                  Originally posted by PuckPoltergeist View Post

                  @michael
                  That's wrong! It's not about spinlocks from Linux. The developer who blamed the kernel has written his own spinlock in userspace without necessary knowledge. He did his own locking wrong and was measuring the wrong numbers. Spinlocks from Linux kernel are completely out of scope here.
                  Michael Is it possible for you to edit the article to reflect this? It seems to be very important.

                  Comment


                  • #89
                    Originally posted by brainlet_pederson
                    He doesn't care. Gotta get those flamebait clicks.
                    It's probably more a combination of the fact that he's really busy and doesn't really have a strong incentive to do it.

                    If he were actively trying to misrepresent the facts, for the sake of clickbait, it would quite obvious. This seems like an honest error, but I haven't seen details on this level frequently corrected, here.

                    Comment


                    • #90
                      Originally posted by Creak View Post
                      Whether you like it or not, Linus is quite rude in his answer. No matter how right or wrong the Google dev is, Linus sets an example of how you will be received if you tell something on the mailing list that you believe is right. This doesn't set a friendly environment and doesn't encourage new comers to share their issues, even if theirs might be real ones.
                      How could you arrive at such an interpretation? When someone speaks with great confidence over things they have no clue about, should they be patted on the back? Especially when they go and insult and attack Linus's kernel. He was really polite considering the context.

                      Comment

                      Working...
                      X