Announcement

Collapse
No announcement yet.

GCC Compiler Tests At A Variety Of Optimization Levels Using Clear Linux

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

  • GCC Compiler Tests At A Variety Of Optimization Levels Using Clear Linux

    Phoronix: GCC Compiler Tests At A Variety Of Optimization Levels Using Clear Linux

    For those curious about the impact of GCC compiler optimization levels, a variety of benchmarks were carried out using GCC 6.3 on Intel's Clear Linux platform.

    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
    This is great info.

    Would be really interesting to see how the Intel compiler (and maybe others?) compares as well. Back in the day it had a rather large advantage compared to GCC regardless of optimization settings, but who knows if that is still the case.

    Comment


    • #3
      It's too bad AMD isn't developing Open64 anymore. It was the closest thing we ever had to a real replacement to GCC. It never was able to build everything, but it was closer and closer a while back. And it definitely outperformed GCC on the performance of generated binaries.

      Comment


      • #4
        For SciMark, "-O3 -march=native" slower than "-O3" ? Probably not the only edge case.

        Comment


        • #5
          Originally posted by indepe View Post
          For SciMark, "-O3 -march=native" slower than "-O3" ? Probably not the only edge case.
          It most commonly happens if -march=native is added to what aren't the most optimal flags.

          Interesting that I had different results with some of the programs where I did find improvements using -march=native, but none appeared in these tests.

          Comment


          • #6
            it seems (from my limited understanding) that "-O2 -march=native" seems the be the optimal setting with the potential flaws of -O3 not making it worth while for such a small performance jump

            Comment


            • #7
              Thank you for adding the time and size charts as well!

              Comment


              • #8
                Nice seeing -Og doing its job

                Comment


                • #9
                  Originally posted by carewolf View Post
                  Nice seeing -Og doing its job
                  Yes, I just thought I should look into using it more often for dev builds.

                  Comment


                  • #10
                    Interesting, but I think it also depends on the architecture you build for. These may also vary. -Os is said to be sometimes a little better for older CPUs with small caches. For my normal machines I guess I'll stay with -O2 wich is said to be stable, while -O3 is suspected to occasionally produce hiccups on complex code.
                    Stop TCPA, stupid software patents and corrupt politicians!

                    Comment

                    Working...
                    X