AMD AOCC 3.2 vs. GCC vs. Clang Compiler Performance On Zen 3
First up with the QuantLib quantitative finance modeling library, the performance wasn't all that different besides the GCC 10.3 release performing much slower than the rest. It was not until GCC 11 that the AMD Zen 3 optimizations were introduced for the GNU Compiler Collection. AOCC 3.2 did come out technically in front for this benchmark but with no meaningful difference against the newer GCC and Clang releases.
Meanwhile with the Etcpak ETC texture compressor, LLVM Clang upstream was performing better than the AOCC downstream.
Or with WebP image encoding, AOCC was on-par with the upstream LLVM Clang releases.
For many of the tests there were very slim differences in the results. GCC 12.0 did rise to the occasion of delivering slightly better Zstd compression performance.