GCC 8/9 vs. LLVM Clang 7/8 Compiler Performance On AArch64

Written by Michael Larabel in Software on 12 February 2019. Page 2 of 4. 20 Comments

With the MAFFT scientific application, the Clang performance barely inched ahead of GCC 9, which in turn was coming in slightly faster than GCC 8.2 but with some deviation in the results.

Like with the POWER9 tests, when running the CacheBench micro-benchmark on Clang with this 64-bit ARM server its performance was much better than GCC 8/9.

For the SciMark micro-benchmarks, GCC ended up being about 16% faster than Clang.

Clang did stay neck-and-neck with GCC in the GraphicsMagick imaging tasks that weren't too heavily threaded with OpenMP.

Meanwhile with the Himeno pressure solver benchmark, the binaries built by GCC were over 32% faster than Clang.

Related Articles