AMD's FX-8150 Bulldozer Benefits From New Compilers, Tuning
Without AMD Bulldozer-specific optimizations, LLVM/Clang was the compiler producing the fastest binaries for the eight-core FX-8150 processor. LLVM/Clang in its stock configuration was noticeably faster than GCC and Open64. However, when taking advantage of the bdver1 support, GCC 4.7.0 quickly turned into being the compiler with the fastest Bulldozer binaries with LLVM seeming to not benefit from the bdver1 option. The C-Ray multi-threaded ray-tracing time dropped in half with Bulldozer optimizations over the generic K8 optimizations while the Open64 performance improved by about 20%.