Talk:AES instruction set

Software supporting AES instruction set

Would anyone object to limiting software to software libraries? It seems like the list will grow until it is unmanageable if the only restriction is software. If the restriction is software libraries, the growth will be checked, and it will provide useful information for programmers who need a list of supporting libraries.

JW Noloader (talk) 02:00, 13 August 2010 (UTC)[reply]

Windows' Bitlocker should also be added to the list. 72.80.29.25 (talk) 14:55, 12 July 2011 (UTC)[reply]

Security

There is no discussion about security implications. In the non-hw assisted case the CPU has a hard time figuring out what is your key. In this case, you're essentially giving a black box your key, AND telling it the semantics too "this here is my key".

How can you be sure there isn't (for example) a built-in SRAM cache which stores the most-recently used 500 keys?

94.254.76.147 (talk) 09:37, 2 June 2011 (UTC)[reply]

Comparison with a Pentium 4

I think this comparison is...questionable. Especially considering, that a Pentium 4 is well known to have the worst IPC of its generation, and was out of date by three generations by the time AES-NI became available, including one major and one minor architectural change. So looking at the cycles per byte of the worst offender of long pipelines, with the Netburst architecture, and the cycles per byte using optimized, dedicated logic is very strange. A comparison with a pure software implementation on the same platform would be interesting, or maybe with a previous platform, but a comparison with a P4 appears rather useless. 193.49.124.107 (talk) 08:27, 17 August 2011 (UTC)[reply]