Announcement

Collapse
No announcement yet.

GCC 4.9 Shows Little Change For AMD's Kaveri

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

  • GCC 4.9 Shows Little Change For AMD's Kaveri

    Phoronix: GCC 4.9 Shows Little Change For AMD's Kaveri

    Our latest Linux benchmarks of AMD's new "Kaveri" APU with Steamroller processor cores is comparing the GCC 4.8.2 compiler performance to the very latest GCC 4.9 compiler snapshot to see how the performance is fairing for this next Free Software Foundation compiler release due out within the next few months.

    http://www.phoronix.com/vr.php?view=19783

  • #2
    "-Ofast" will almost certainly give you faster binaries. You just need to be sure your code is "fast-math" safe.

    Comment


    • #3
      Originally posted by s_j_newbury View Post
      "-Ofast" will almost certainly give you faster binaries. You just need to be sure your code is "fast-math" safe.
      What does this mean and how does one accomplish this? I'm just trying to learn here!

      Comment


      • #4
        Originally posted by Tgui View Post
        What does this mean and how does one accomplish this? I'm just trying to learn here!
        ffast-math makes some assuptions about float math which can lead to faster code and loss of precision. see the gcc manual about it.

        Comment


        • #5
          Originally posted by Tgui View Post
          What does this mean and how does one accomplish this? I'm just trying to learn here!
          -O3 vs -Ofast

          @@ -23,7 +23,7 @@
          -fcrossjumping [enabled]
          -fcse-follow-jumps [enabled]
          -fcx-fortran-rules [disabled]
          - -fcx-limited-range [disabled]
          + -fcx-limited-range [enabled]
          -fdata-sections [disabled]
          -fdce [enabled]
          -fdefer-pop [enabled]
          @@ -35,7 +35,7 @@
          -fearly-inlining [enabled]
          -fexceptions [disabled]
          -fexpensive-optimizations [enabled]
          - -ffinite-math-only [disabled]
          + -ffinite-math-only [enabled]
          -ffloat-store [disabled]
          -fforward-propagate [enabled]
          -fgcse [enabled]
          @@ -73,7 +73,7 @@
          -floop-nest-optimize [disabled]
          -floop-parallelize-all [disabled]
          -floop-strip-mine [disabled]
          - -fmath-errno [enabled]
          + -fmath-errno [disabled]
          -fmerge-all-constants [disabled]
          -fmerge-constants [enabled]
          -fmodulo-sched [disabled]
          @@ -127,7 +127,7 @@
          -fshort-wchar [disabled]
          -fshrink-wrap [enabled]
          -fsignaling-nans [disabled]
          - -fsigned-zeros [enabled]
          + -fsigned-zeros [disabled]
          -fsingle-precision-constant [disabled]
          -fsplit-ivs-in-unroller [enabled]
          -fsplit-wide-types [enabled]
          @@ -136,7 +136,7 @@
          -fthread-jumps [enabled]
          -fno-threadsafe-statics [enabled]
          -ftoplevel-reorder [enabled]
          - -ftrapping-math [enabled]
          + -ftrapping-math [disabled]
          -ftrapv [disabled]
          -ftree-bit-ccp [enabled]
          -ftree-builtin-call-dce [enabled]
          @@ -180,7 +180,7 @@
          -funroll-all-loops [disabled]
          -funroll-loops [disabled]
          -funsafe-loop-optimizations [disabled]
          - -funsafe-math-optimizations [disabled]
          + -funsafe-math-optimizations [enabled]
          -funswitch-loops [enabled]
          -funwind-tables [disabled]
          -fvar-tracking [enabled]

          Comment


          • #6
            Originally posted by Tgui View Post
            What does this mean and how does one accomplish this? I'm just trying to learn here!
            It will violate the IEEE standard in subtle ways that usually doesn't matter. The same setting on the intel compiler is enabled by default.
            Last edited by carewolf; 02-01-2014, 07:39 AM.

            Comment

            Working...
            X