AMD AOCC 2.2 Helping Squeeze Extra Performance Out Of AMD EPYC 7002 "Rome" CPUs
In some of the benchmarks there were slight advantages to AOCC 2.2 compared to GCC 10.2, but in some of those cases the performance was simply comparable to upstream LLVM Clang 10 for which AOCC is based.
But in other test cases like MrBayes molecular biology or Zstd compression, AOCC 2.2 was faster than GCC 10.2 which in turn was faster than Clang 10.