GCC 11 vs. LLVM Clang 12 Compilers On The AMD EPYC 7763

GCC was still favored for the LibRaw RAW image decoder library.

Clang meanwhile produced slightly faster binaries for the TSCP chess benchmark.

GraphicsMagick with its OpenMP threading tended to perform the best on GCC 11 but for some sub-tests Clang 12 came out ahead.

The dav1d AV1 video decode results were quite close with the exception of decoding 10-bit AV1 content where the GCC compiler yielded a faster decoder. This is largely to be expected with dav1d relying on hand-tuned x86_64 Assembly but currently lacking such 10-bit optimized code for x86_64.

The AOM-AV1 and Intel SVT encoders tended to be quite close between compilers, again with most performance sensitive video encode/decode code being hand-tuned Assembly these days, but in some of the Intel test cases Clang did seem to show an advantage at least for this Zen 3 server.