Announcement

Collapse
No announcement yet.

Benchmarks Of GCC 4.5.0 Compiler Performance

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

  • Benchmarks Of GCC 4.5.0 Compiler Performance

    Phoronix: Benchmarks Of GCC 4.5.0 Compiler Performance

    Last week GCC 4.5.0 entered the world with improvements to the experimental C++0x support, Graphite-powered automatic parallelization support, compatibility with new ARM processors, Intel Atom and AMD Orochi optimizations, link-time optimization, and GCC plug-in support. Over the weekend we decided to benchmark this major update to the GNU Compiler Collection to see how its performance compares to that of GCC 4.3 and 4.4.

    http://www.phoronix.com/vr.php?view=14808

  • #2
    It would be better to build entire system (Gentoo provides great framework for that) with either 4.3, 4.4 and 4.5 and THEN test performence of the output code for the whole system built with exact compiler.

    Comment


    • #3
      Originally posted by vermaden View Post
      It would be better to build entire system (Gentoo provides great framework for that) with either 4.3, 4.4 and 4.5 and THEN test performence of the output code for the whole system built with exact compiler.
      Actually it would be more interesting to see how the performance of the resulting applications is affected by the new gcc optimizations. The build process itself is something you only do once in a while, but if the programs run faster its worthwhile the extra initial build time.

      Comment


      • #4
        Originally posted by screemo View Post
        Actually it would be more interesting to see how the performance of the resulting applications is affected by the new gcc optimizations. The build process itself is something you only do once in a while, but if the programs run faster its worthwhile the extra initial build time.
        I meant to build these Gentoo's and THEN compare applications' performance differences, not the compile time or anything like that.

        Comment


        • #5
          Originally posted by vermaden View Post
          I meant to build these Gentoo's and THEN compare applications' performance differences, not the compile time or anything like that.
          Oh sorry, we just meant the same thing then

          Comment


          • #6
            @screemo

            No problem, maybe I wasnt clear enough

            Comment


            • #7
              Between GCC releases performance should not improve that greatly on the same command line arguments. Only in earlier releases, GCC did get improvements but as of today the optimizations strategies are well tested and improved. Probably a better test IMHO was between LLVM and GCC 4.5
              In fact all the tests are fully supported by GCC like OpenMP and autovectorization support.
              Wanna see bigger difference to appear a more interesting article?
              Compile with link time optimization s raytracer, or a fortran scientific system. Not that representative for user's software usage?
              But testing three compilers under the same algorithm/parameters, will make mostly to not get big gains. The latest problem is that even they will bring 10% speedup, this will not affect regular Linux user anyway, as a lot of applications are not performance critical, and are written in python or bash, or for server based case, filesystem and IO affect much more than the small speedup that a compiler brings.

              Comment


              • #8
                Good to see C-Ray regression fixed. Oops, but there's a regression BFE now.

                Comment


                • #9
                  Oops, but there's a regression in BFE now.
                  Sorry, I missed one minute edit window... In bullet physics engine.

                  Comment


                  • #10
                    Very interesting.
                    I'm curious: was "Graphite" enabled in GCC 4.4/4.5?
                    (Does Graphite make a difference?)

                    Comment


                    • #11
                      Intel C

                      It will be very interesting to compare with Intel's C compiler: http://software.intel.com/en-us/intel-compilers/
                      There is evaluation version for 30 days

                      Comment


                      • #12
                        Did you try usig the -flto flag for link time optimization, the -fwhole-program flag for non libraries or -floop-parallelize-all?

                        These are the new features that'll make GCC 4.5 worth while

                        Comment


                        • #13
                          Originally posted by FireBurn View Post
                          Did you try usig the -flto flag for link time optimization, the -fwhole-program flag for non libraries or -floop-parallelize-all?

                          These are the new features that'll make GCC 4.5 worth while
                          Yeah I was going to ask the same. GCC 4.4 also needed -floop-interchange -floop-strip-mine -floop-block to enable graphite which I'd suspect would affect benchmarks versus GCC 4.3

                          Comment


                          • #14
                            I have always wondered how complex compiler are.

                            Are the people that make GCC insanely inteligent math nerds? Ofcourse no pun intended, rather respect.

                            Comment


                            • #15
                              Originally posted by V!NCENT View Post
                              I have always wondered how complex compiler are.

                              Are the people that make GCC insanely inteligent math nerds? Ofcourse no pun intended, rather respect.
                              Insanely intelligent or insanely persistent.... not that there's a provable difference between the two. In either case though, definitely math nerds.

                              And no, they definitely wouldn't take any kind of offense to that.

                              Comment

                              Working...
                              X