GCC vs. LLVM Clang Compiler Performance On The AMD Threadripper 2990WX Benchmarks

The PolyBench-C test profile tends to not be too interesting but in this case of three matrix multiplications, GCC 8.2 was faster than GCC 7.3 and now GCC 9.0 SVN is shaping up to be much faster than GCC8... GCC 8 in particular received a lot of AMD Zen (znver1) performance tuning. The GCC 9.0 performance is now on-par with Clang 6.0.1 in this particular benchmark.

GCC 9.0.0 is also looking good with FFTW and continues to outperform Clang there.

While the TSCP chess benchmark is one of the workloads where Clang continues outperforming GCC.

GCC 9.0 appears to deliver a boost in x265 video encode performance, at least for this Zen+ 2990WX system.

In the OpenMP-threaded GraphicsMagick benchmarks, GCC tends to continue doing better than Clang. At least these days Clang has OpenMP support but it still doesn't seem to be as mature as GCC's implementation.