There was not much difference in the Gcrypt CAMELLIA256-ECB Cipher performance aside from when playing with the loose optimizations levels of -O0 and -O1.
The results from this first round of compiler tuning benchmarks are not terribly surprising. The more aggressive optimizations, well, generally deliver better performance. Worth noting though is that the "bdver1" optimizations were not too beneficial when used alone with these tests / code (the Bulldozer-specific optimizations will be tested in another article in the near future with the bdver1 support in GCC too). Testing some of the other individual Bulldozer/Interlagos optimizations recommended by AMD tended to not be too beneficial for the performance with the select tests used in this article.