AMD AOCC 1.2 Code Compiler Offers Some Performance Benefits For EPYC

Last month AMD released the AOCC 1.2 compiler for Zen systems. This updated version of their branched LLVM/Clang compiler with extra patches/optimizations for Zen CPUs was re-based to the LLVM/Clang 6.0 code-base while also adding in experimental FLANG support for Fortran compilation and various other unlisted changes to their "znver1" patch-set. Here's a look at how the performance compares with AOCC 1.2 to LLVM Clang 6.0 and GCC 7/8 C/C++ compilers.

On the TYAN B8026T70AE24HR server powered by an AMD EPYC 7601, I was running Ubuntu 18.04 LTS x86_64 with the Linux 4.15 kernel and from there tested the release builds of GCC 7.3.0, GCC 8.1.0, LLVM Clang 6.0.0, and AMD AOCC 1.2 for seeing how these latest compiler releases compare. A variety of open-source C/C++ benchmarks were built from source with each compiler while the CFLAGS/CXXFLAGS were set to "-O3 -march=native" for the entire duration of benchmarking.

All of these benchmarks were carried out in a fully-automated and reproducible manner using the open-source Phoronix Test Suite benchmarking software. For those unfamiliar with the AMD Optimizing C/C++ compiler (AOCC), see our past articles and benchmarks on the topic.