Announcement

Collapse
No announcement yet.

GCC 4.9 vs. GCC 5.1 vs. GCC 6.0 SVN Compiler Benchmarks

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

  • GCC 4.9 vs. GCC 5.1 vs. GCC 6.0 SVN Compiler Benchmarks

    Phoronix: GCC 4.9 vs. GCC 5.1 vs. GCC 6.0 SVN Compiler Benchmarks

    Here's some new GCC compiler benchmarks on Linux x86_64 for your viewing pleasure this weekend...

    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
    Displaying standard deviation everywhere would be nice in these tests, because probablly almost all of the tests have such small difference the result is meaningless for comparison?
    Of course in a way that is also a result, it shows not to expect anything great from gcc5.1 if you are only looking for performance numbers.
    Last edited by utack; 07 June 2015, 11:40 AM.

    Comment


    • #3
      Originally posted by utack View Post
      Displaying standard deviation everywhere would be nice in these tests, because probablly almost all of the tests have such small difference the result is meaningless for comparison?
      Of course in a way that is also a result, it shows not to expect anything great from gcc5.1 if you are only looking for performance numbers.
      Did you look at the results? The deviation is reported on every single Phoronix result, always.
      Michael Larabel
      https://www.michaellarabel.com/

      Comment


      • #4
        Michael, After you built each compiler did you then build it again with itself before doing the benchmarks?

        Comment


        • #5
          Originally posted by jsl.com View Post
          Michael, After you built each compiler did you then build it again with itself before doing the benchmarks?
          GCC already does that. It first builds a "small" version of itself, and then recompiles the codebase with that one.

          Comment


          • #6
            Originally posted by mlau View Post

            GCC already does that. It first builds a "small" version of itself, and then recompiles the codebase with that one.
            But even if this wasn't the case: the choice of the compiler for compiling the compiler won't change the performance of the binaries at the very end. Only the performance of the compiler while creating them.

            Comment


            • #7
              Originally posted by utack View Post
              Displaying standard deviation everywhere would be nice in these tests, because probablly almost all of the tests have such small difference the result is meaningless for comparison?
              Of course in a way that is also a result, it shows not to expect anything great from gcc5.1 if you are only looking for performance numbers.
              Gcc 5 has several significant optimization improvements. They may just not be triggered in these few tests.

              Comment

              Working...
              X