AMD Piledriver/Trinity A10-5800K Compiler Tuning

Himeno wasn't able to take advantage of the more modern AMD CPU features.

When targeting the modern CPUs with broader instruction set coverage, the compilation process of code takes slightly longer.

The C-Ray ray-tracing benchmark saw improvements out of targeting Barcelona and then bdver1, but similar to the other software tested thus far it didn't see any measurable performance improvement out of "bdver2" with the FMA3, BMI, TBM, and F16C instruction set additions.