Announcement

Collapse
No announcement yet.

LLVM's LLD Linker Gets Faster Performance (Parallelized ICF)

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

  • LLVM's LLD Linker Gets Faster Performance (Parallelized ICF)

    Phoronix: LLVM's LLD Linker Gets Faster Performance (Parallelized ICF)

    As pointed out by this week's LLVM Weekly, the LLVM Linker (LLD) received a rather nice performance optimization this past week...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    "When the parallel ICF was ported to COFF, it's a big win too with LLD formerly taking 11.73 seconds to link Clang while now it's down to 6.94 seconds while Microsoft MSVC link takes 83 seconds."
    This is for those kind of reasons I can't understand anymore how can someone sincerely like Microsoft's C++ toolchain.

    Comment


    • #3
      Can it be used as a drop in replacement for ld.bfd or ld.gold? Even ld.gold still struggles with some packages (granted its a very low number now)

      Comment


      • #4
        Considering this is parallel. It is not an additional thing on top of --threads, it depends on --threads being enabled. Which also means you should compare with gold running threaded.

        Comment


        • #5
          Originally posted by AsuMagic View Post
          "When the parallel ICF was ported to COFF, it's a big win too with LLD formerly taking 11.73 seconds to link Clang while now it's down to 6.94 seconds while Microsoft MSVC link takes 83 seconds."
          This is for those kind of reasons I can't understand anymore how can someone sincerely like Microsoft's C++ toolchain.
          I don't think anyone has ever loved MS's toolchain, so much as their IDE's.

          Comment


          • #6
            Originally posted by carewolf View Post
            Considering this is parallel. It is not an additional thing on top of --threads, it depends on --threads being enabled. Which also means you should compare with gold running threaded.
            Right, because I'm sure these seasoned GCC developers are unaware of this.

            Of course the Google team demonstrating LLD Parallel are the same company that developed ld-gold that you are worried about getting embarrassed.

            Comment


            • #7
              Performance is one thing, but can I throw away the other linkers on my windows box and keep lld?

              Comment


              • #8
                Originally posted by Marc Driftmeyer View Post

                Right, because I'm sure these seasoned GCC developers are unaware of this.

                Of course the Google team demonstrating LLD Parallel are the same company that developed ld-gold that you are worried about getting embarrassed.
                Of course they are aware of this, this is why the numbers so cherry-picked and poorly specified. I was pointing it out for the story because it didn't make that clear, and Michal is not GCC or LLD developer in case you are confused.

                Comment

                Working...
                X