It's basically a limitation of managed code (not to rule out that our
implementation might yet be improved).
You might consider experimenting with unsafe code in
EncryptBlock/DecryptBlock to see if there are array bounds checks that
can be eliminated, but I don't really have any idea what sort of
improvement that's likely to give, if any.
Pete.
On 18/01/2013 9:00 PM, Ray Kelly wrote:
I am decrypting a CFB8/No Padding Stream. BouncyCastle is about 80%
slower that the native decryption (I cant use native Rijndael in my
project for various reasons). I ran a test decrypting 30k bytes in a
single ProcessBytes call. I compared the native
Rijndael(TransformBlock) to the AesFastEngine and BouncyCastle
Rijndael. Here are the results:
Native Rijndael: 21ms
AesFastEngine: 34ms
BC Rijndael: 288ms
Native Rijndael: 19ms
AesFastEngine: 31ms
BC Rijndael: 288ms
Native Rijndael: 19ms
AesFastEngine: 31ms
BC Rijndael: 287ms
Is there another engine I should use or something I am doing wrong?
Thanks for your help,
Ray
