Announcement

Collapse
No announcement yet.

LLVM's Clang Is Finally The FreeBSD x86 Compiler

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

  • #16
    Originally posted by blackout23 View Post
    Since when is that something that makes an OS good? That it adopts something (which isn't even a clear improvement) for whatever reason.
    It did not adapt systemd with that logic it would make it a bad OS. There is no Gnome 3 for BSD they're still at 2.3.x something. What does that tell us? Nothing.
    rare found wise words

    Comment


    • #17
      Why is there often such hate when a *BSD article gets discussed?
      Can't we all just love each other?

      Comment


      • #18
        It would be nice...

        It would be lovely to see some benchmarks:
        - comparing FreeBSD with Clang vs FreeBSD with outdated GCC.
        - comparing FreeBSD with Clang vs FreeBSD with latest GCC.
        - comparing FreeBSD with Clang vs Linux (but a test that could really say something meaningful about the kernels, not just another silly UFS vs ExtFS).

        Comment


        • #19
          Originally posted by Cthulhux View Post
          I meant: They're faster adapting the CLANG compiler. But after all, I don't know "the only true" benchmarks.
          Well they made Clang to get rid of GCC in the first place. Linux wants to stay with GCC. There aren't "the only true benchmarks" obviously, but from the ones here on phoronix I've been seeing that GCC is mostly faster than Clang, but I recall specific cases where Clang was faster, but hose were a minority.

          Originally posted by vertexSymphony View Post
          Cite those benchmarks ... I hope those benchmarks are contextualized ( BSD benchmarks comparing the BSD situation: GCC4.2 vs latest Clang )
          About OS performance ... well, compiler does its part, but thinking it's the only thing is just nonsense.

          In any case Paws Up for the FreeBSD people for switching to a wayy better c/c++ stack ( this also includes libc++, libcxxrt and some work with binutils kind of tools ) and for putting work into this.
          It would interesting to do some benchmarks, but with the new stack in place ... I kinda doubt that we would loose performance against an raging old version of GCC, but I would like to see how the other improvements interact with that (because it wasn't just Clang)
          I agree, in that any opensource "competition" to GCC should only be helpful to its users too. I recall the comparison was with GCC on Linux? Whatever ..
          Won't comment much on "better stack", I don't have much knowledge on whether it would be better or not.
          I'd say for benchmarks on BSD with new GCC too, as well as Clang etc on Linux too, if they are done.

          Comment


          • #20
            Originally posted by Cthulhux View Post
            I meant: They're faster adapting the CLANG compiler.
            There's been no attempt from the Linux kernel devs to 'adapt' to Clang/LLVM, what little progress we've seen in that area are from third-party efforts. Linux has no need to switch compiler toolchain and I've seen no indication that there is a wish to do so from the kernel devs.

            As for FreeBSD switching to Clang/LLVM, that was a very simple choice even though it took alot of effort. As others pointed out they were stuck at GCC 4.2 which is a ~5 year old compiler, the reason they were stuck was because GPLv3 was not acceptable in accordance with the goals of FreeBSD's project aswell as that of their corporate investors. Add to this the fact that Clang/LLVM was using permissive licencing which of course gels much better with FreeBSD's ideology.

            As far as I can tell, having to rely on GPL licenced software for such a core component as the compiler toolchain has also been a thorn in their side. Beyond this, I'm sure they also like the idea of a more modern compiler toolchain, but really I'm pretty sure the first reasons I listed took preference. Anyway the switch has taken a long time but now it seems it is finally done for the x86/x64 architectures.

            edit: and please let's not have this thread deteriorate into a BSD vs GPL shitfest...

            Comment


            • #21
              How do you benchmark a kernel anyway ? Most of the benchmarks on Phoronix are testing programs.
              What will a kernel benchmark include ? latency ? throughput ? memory consumption ? better scheduling ?

              Comment


              • #22
                Originally posted by mayankleoboy1 View Post
                How do you benchmark a kernel anyway ? Most of the benchmarks on Phoronix are testing programs.
                What will a kernel benchmark include ? latency ? throughput ? memory consumption ? better scheduling ?
                Probably like this: http://www.cs.virginia.edu/~kc5dm/projects/osperf.pdf

                Comment


                • #23
                  Fuck this idiotic crappy Clang!!!!!!!!!! Fucking shit Clang doomed little chicken FreeBSD to death. Lol!!!!!!
                  Last edited by China_Guy; 11-08-2012, 02:28 AM.

                  Comment


                  • #24
                    Originally posted by BitRot View Post
                    Altough some of the methods might make sense the paper itself really doesn't tell much if anything about the kernel performance. They tested the operating systems with one very specific and uncommon set of hardware with single core processor. They also tested only very few factors that affect kernel performance. The conclusion is in all likelhood false and at very least dated.

                    Comment


                    • #25
                      Originally posted by blackout23 View Post
                      There is no Gnome 3 for BSD they're still at 2.3.x something.
                      That's actually wrong.
                      OpenBSD is shipping Gnome 3 since version 5.1.

                      Comment


                      • #26
                        Originally posted by BitRot View Post
                        Interesting... altought I think far from ideal, it is much more relevant than Phoronix tests.

                        Comment

                        Working...
                        X