Announcement

Collapse
No announcement yet.

AMD Ryzen Threadripper Compiler Tuning/Optimization Benchmarks With GCC 9, PGO

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

  • AMD Ryzen Threadripper Compiler Tuning/Optimization Benchmarks With GCC 9, PGO

    Phoronix: AMD Ryzen Threadripper Compiler Tuning/Optimization Benchmarks With GCC 9, PGO

    For those interested in compiler optimization/tuning with AMD Ryzen Threadripper hardware, here are some follow-up benchmarks to Tuesday's GCC 9 vs. Clang 8 C/C++ Compiler Performance On AMD Threadripper, Intel Core i9...

    http://www.phoronix.com/scan.php?pag...-Optimized-Run

  • #2
    Benchmarking LTO and PGO without benchmarking them together is like benchmarking -march=native without -O3 or like comparing fries to burgers without including the combined taste or like tooth paste without tooth brush or ...
    Last edited by sdack; 05-16-2019, 07:33 AM.

    Comment


    • #3
      edit: my bad, this was on the 8086K box, not the 2700X. Yesterday on the (...) box, I built gcc9 with -02 vs -03 profiled, and then gcc recompiled itself about 5-6% faster. I do need to retry this on the 2700X for reasons.
      Last edited by xorbe; 05-16-2019, 03:20 PM.

      Comment


      • #4
        Originally posted by xorbe View Post
        Yesterday on a 2700X, I built gcc9 with -02 vs -03 profiled, and then gcc recompiled itself about 5-6% faster.
        I see an ambiguity in your post. Do you mean that -O2 is faster or -O3-profiled is faster?

        Where you able to verify that the 5-6% speedup translates to non-gcc source codes (that is: to C/C++ codes that aren't part of gcc itself, such as glibc/qt/chromium/mesa/vim/linux-kernel/...)?

        Comment

        Working...
        X