Re: AES counter mode

On Sun, Jun 2, 2013 at 2:57 PM, Peter Gutmann <[email protected]> wrote:
> Kostantinos Koukopoulos <[email protected]> writes:
>>Also am I wrong to say that I cannot encrypt the plaintext blocks in parallel
>>since I need to have already encrypted all previous blocks to encrypt the
>>next one?
>
> Yes.  OTOH unless you're using specialised crypto hardware you can't do much
> in parallel for CTR either, you're still only doing on ECB block at a time.
> Intel has published whitepapers showing impressive results from
> parallelisation with e.g. 6 cores and 12 threads, but it's unclear whether you
> could ever do that in a real-world environment, you'd need to grab control of
> every thread on every core for your own use which I suspect no OS would let
> you do.

Understood, although I think "parallel" was a bit of an overstatement.
 What I basically had in mind was encrypting and decrypting blocks in
an arbitrary order, i.e. not necessarily in a monotonic sequence. I
understand that's not an obviously useful procedure, but it's
interesting to note that AES CTR could do that.

Thanks again!
Konstantinos

_______________________________________________
Cryptlib mailing list
[email protected] via Mail: [email protected]
Archive: ftp://ftp.franken.de/pub/crypt/cryptlib/archives/
http://news.gmane.org/gmane.comp.encryption.cryptlib
Posts from non-subscribed addresses are blocked to prevent spam, please
subscribe in order to post messages.