GCC 4.6 Compiler Performance With AVX On Sandy Bridge
With MrBayes, another molecular biology test, enabling the Core i7 optimizations led to a dramatic speed-up that more than doubled its performance. Utilizing the corei7-avx support though had not helped and when enabling the AVX floating-point math we see again another test that is currently impaired by this flag.
Using any Core 2 / Core i7 optimizations with GCC really hurt HMMer's performance, which is another scientific application.
The FLAC audio encoding performance marginally got faster with the newer GCC releases and when using the different CPU flags.