AMD AOCC 1.1 Shows Compiler Improvements vs. GCC vs. Clang

In some benchmarks the AOCC 1.1 compiler was slower than LLVM Clang upstream but in line between GCC 7.2 and GCC 8.0.1.

With SciMark2's Monte Carlo benchmark, AOCC 1.1 was the fastest compiler tested.

But with the FFT and matrix factorization micro-benchmarks for instance it was slightly slower than the rest.

While in the case of the Jacobi Successive Over-Relaxation benchmark, AOCC 1.1 was faster than the tested upstream Clang compilers but behind that of GCC 7.2/8.0.1.