Request for Review: AESCrypt.init performance JDK change (original) (raw)

Deneau, Tom tom.deneau at amd.com
Mon Jul 16 13:08:45 PDT 2012

Please review the following webrev which improves performance for the com/sun/crypto/provider when the algorithm is AES for some kinds of Cipher usage patterns.

http://cr.openjdk.java.net/~tdeneau/aescrypt-init-performance/webrev.01/

I do not have a bug number for this change but a description would be something like the following:

Consider an application which holds on to a Cipher object for AES and keeps the same key but calls Cipher.init to change the Cipher between encryption and decryption. The current code in AESCrypt.java recomputes the expanded key (which is an expensive operation) with each call to init.

The proposed code saves the key and, if it has not changed, does not need to recompute on subsequent calls to Cipher.init. The proposed code also incorporates into the key calculation the final step which was the expansion of the key to a 1-dimensional array of ints. This step used to be done in the routine setSubKey.

-- Tom Deneau

More information about the jdk7u-dev mailing list