GCC 8.0 vs. LLVM Clang 6.0 On AMD EPYC
Znver1 was also of some benefit in the GraphicsMagick benchmarking, but Clang was trailing behind GCC 8.0 with its OpenMP implementation not being as strong.
In the Himeno pressure solver benchmark, znver1 targeting regressed GCC 8's performance while Clang 6.0 saw a slight improvement in performance and was in first place for this test.
Ebizzy was another test example where GCC 8 didn't see much out of znver1 while Clang 6.0 did. GCC 8 has seen znver1 improvements, but it looks like it (as well as LLVM/Clang) could still see more optimization work for these Zen CPU cores.
With the C-Ray multi-threaded ray-tracer the scenario was opposite where znver1 really helped GCC 8 but was of minimal help to Clang 6.0.
Znver1 really helped Clang with FLAC audio encoding to secure a first place finish.
But with LAME MP3 encoding, znver1 tuning regressed Clang's performance.
Those wishing to dig through more of this compiler benchmark data can continue to OpenBenchmarking.org for additional benchmark results.
If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.