GCC vs. LLVM Clang vs. AOCC Compiler Benchmarks On The AMD EPYC 7742 2P Linux Server
The FFTW performance of the LLVM Clang compiler continues trailing GCC, which we've seen on other x86_64 hardware as well. So to little surprise, AOCC 2.0 with being derived from LLVM Clang is also slower here than GCC. Interestingly here GCC 9.1 came out slightly ahead of the GCC 10 compiler state.
But when it comes to the TSCP chess benchmark, Clang does very well. The upstream Clang 9.0 SVN code outperformed AOCC 2.0 derived from Clang 8, but it still came out measurably ahead of the two GCC releases tested.
Clang 9.0 and AOCC 2.0 were tied for the top spot with the John The Ripper crypto benchmarks, which were about 25% faster than GCC.
AOCC 2.0 did yield a slight performance advantage when it came to the x264 video encoder performance.
The LLVM-based compilers also were managing an extra frame per second with the x265 compiler on this AMD EPYC Rome server.