AMD AOCC 1.2 Code Compiler Offers Some Performance Benefits For EPYC
With the Fhourstones benchmark, LLVM Clang 6.0 upstream was roughly the same speed as GCC 8.1, but the AOCC 1.2 release ended up being 3.5% faster than Clang and about 4.5% faster than GCC 8.1.
In SciMark's Sparse Matrix Multiply run, AOCC 1.2 was almost 4% faster than GCC 8.1.0.
While in the other SciMark benchmark runs, AOCC 1.2 was roughly the same speed as Clang 6.0 for which it is based and in most instances comparable figures to GCC 8.1.
Both Clang 6.0 and AOCC 1.2 tended to perform competitively with the mature GCC compiler stack in many of these C/C++ open-source Linux benchmarks.