> Just as an FYI my collegue and I developed an AES-CTR mode cipher for
> OpenSSH a while ago. There was some discussion about it here a few years
> ago. Anyway, if you would like to see the process we used you can get
> the multithreaded AES-CTR mode cipher from
> http://www.psc.edu/networking/projects/hpn-ssh

On [yet another] related note. Just committed AESNI-CBC-SHA1 "stitched"
cipher/hmac combo can be used in SSH context too. Its single-threaded
performance is approximately same as single-threaded AESNI-CTR + SHA1,
even though CBC encrypt is not parallelizeable. Of course one can
multi-thread AESNI-CTR, but one should consider resulting improvement in
proportion to SHA1. I mean AESNI-CTR is ~4 times faster than SHA1, so
that by making AESNI-CTR run on e.g. two cores, overall improvement
can't be more than 11%. Besides, even single-threaded it's faster than
1Gbps. Either AESNI-CBC-SHA1 "stitch" or AESNI-CTR + SHA1 that is. For
reference, it makes no sense to "stitch" AESNI-CTR with SHA1, because it
won't result in any [significant] gain.
______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       [email protected]
Automated List Manager                           [email protected]

Reply via email to