AMD AOCC 2.3 Squeezing Out Extra Performance For EPYC Over GCC 10, Clang 11

For some of the tests AOCC 2.3 had some modest uplift...

Intel's oneDNN neural network library was one of the areas where AOCC 2.3 was helping squeeze extra performance out of the EPYC 7502 processor.

But not all of the tuning paid off, as in a few tests the AOCC performance was in last -- in some cases at least due to LLVm/Clang traditionally performing slower than GCC in those tests.