For some reason, the MSVC 6 compiler likes (unsigned int)(byte)x better than x & 0xff, and is able to optimize code utilizing the former construct more.
On Mon, May 19, 2003 at 09:01:23PM -0500, Cheng Huang wrote: > I have a "strange" observation when testing crypto++ package. The > throughput of AES encryption/decryption using RELEASE version of > crypto++ is about twice faster than the "Optimized C code v3.0" from > link http://www.esat.kuleuven.ac.be/~rijmen/rijndael/, although two > implementations are almost the same. Can anybody explain that to me a > little bit? Thanks. > > -- Cheng >
