GCC 8/9 vs. LLVM Clang 7/8 Compiler Performance On POWER9 With The Raptor Talos II
GCC and in particular GCC 9.0.1 picked up a few more wins when it came to the HINT CPU benchmark.
The Apache web server performance was roughly the same between the tested compilers.
If taking the geometric mean of all the compiler benchmarks carried out where the tests ran on all four compilers, GCC handedly comes out ahead of LLVM Clang on IBM POWER9. This outcome was quite interesting since as shown by the massive Linux x86_64 compiler comparison posted earlier this week on a wide range of x86 64-bit hardware, when it comes to the performance at large Clang was effectively at performance parity with GCC. There was virtually no difference in the overall x86_64 performance but on the POWER9 front, the GNU Compiler Collection averages out to roughly 10% faster. Though it's not entirely surprising with GCC's Power architecture support having been around much longer than LLVM/Clang and seemingly more focus by Clang on x86/ARM architectures. Of some concern though were a few GCC9 performance regressions that led its geometric mean to be lower than GCC 8.2 stable, with each being built in the release mode, while the Clang performance was largely stable between 7.0.1 stable and 8.0-rc.
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.