Announcement

Collapse
No announcement yet.

Linux 3.15 Can Almost Be Compiled Under LLVM's Clang

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

  • Linux 3.15 Can Almost Be Compiled Under LLVM's Clang

    Phoronix: Linux 3.15 Can Almost Be Compiled Under LLVM's Clang

    A few hours ago I wrote about the most interesting features for the Linux 3.15 kernel from my perspective as it didn't look like anything else interesting would be introduced this late in the merge window before the imminent 3.15-rc1. However, this time I've been happily proven wrong with Clang patches being added to the Linux 3.15 kernel...

    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
    and next...?

    I am curious what the developes will do once LLVMLinux enters maintenance mode (all needed patches upstream and LLVM officially supported). I know that one of the earlier LLVM porters later tried Path64 or or one of its forks. There is also the old BSD challenge to use PCC or the more relevant challenge to revive the ability to use TCC (and by extension in the future QCC). Perhaps someone will go all out crazy and try OpenWatcom

    Using proprietary compilers like SunCC or ICC (linux DNA) is less interesting and less likely to ever go upstream. On the other hand, if Linux ever got the DDEkit integrated upstream so that parts of the kernel could be built independently as drivers for different OSes (like the NetBSD kernel can), testing with MSVC and its DDK would make sense (especially file system drivers for NT but also legacy hardware support).

    Comment


    • #3
      Clang is widely accepted as being faster than GCC at compiling code
      Benchmarks usually say the inverse, right on your own site.

      Comment


      • #4
        Originally posted by Calinou View Post
        Benchmarks usually say the inverse, right on your own site.
        Really? Where? In almost all (if not all) benchmarks on this site clang is faster.

        Comment


        • #5
          Originally posted by Szzz View Post
          Really? Where? In almost all (if not all) benchmarks on this site clang is faster.
          I think he is talking about the (useless -O3 -march=native) runtime performance of the code. Clang is always faster in compiling code.

          Comment


          • #6
            Originally posted by Calinou View Post
            Benchmarks usually say the inverse, right on your own site.
            Learn to read Calinou...

            Clang is widely accepted at being much faster to COMPILE code.

            When it comes to the performance of said code ONCE COMPILED Clang is typically very competitive with GCC, except in cases of OpenMP.
            All opinions are my own not those of my employer if you know who they are.

            Comment


            • #7
              Originally posted by -MacNuke- View Post
              I think he is talking about the (useless -O3 -march=native) runtime performance of the code. Clang is always faster in compiling code.
              Run-time performance is "useless" now?

              Tsk, kids these days...

              Comment


              • #8
                --march=native (and usually -O3) compiled code is useless unless you're running gentoo or the like (or compiled the program for your own special purpose), since in most cases (in most distributions) you'll never see that performance yourself.

                Not to say it's not good to know that some compiler or another can create fast performing binaries if you need it to.

                Comment


                • #9
                  Originally posted by dee. View Post
                  Run-time performance is "useless" now?

                  Tsk, kids these days...
                  What the other guy said. Gentoo and Arch (and derivs) are the only two distros I know where its relatively easy to recompile all packages on your system-- eg: built into package manager. Gentoo its the default, Arch has yaourt. No one on any other distro is going to see -O3 -march=native across the board for all users. Most likely they are getting -O2 and architecture set to either generic or MAYBE, just MAYBE, core2duo.
                  All opinions are my own not those of my employer if you know who they are.

                  Comment


                  • #10
                    Originally posted by Ericg View Post
                    Learn to read Calinou...

                    Clang is widely accepted at being much faster to COMPILE code.

                    When it comes to the performance of said code ONCE COMPILED Clang is typically very competitive with GCC, except in cases of OpenMP.
                    Clang/LLVM produce binary that is slower in 99% of arbitrary sources, except in carefully selected microbenchmarks. It widely produce fast code on Apple site and Phoronic site. Other narrow part of world for some reason use ugly slow GCC.

                    Clang optimization levels like -O2 not equal GCC -O2, as it tweaked to use less optimisation passes to claim faster compiling, that is pure BS as same time it just not do some oprimisations.

                    Comment

                    Working...
                    X