[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <4d663a3ea9e637ed096698fc18cff7ea@smtp.hushmail.com>
Date: Thu, 6 Jun 2013 15:45:43 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: [pdf] Parallelizing MD5 on Intel Architectures
On 6 Jun, 2013, at 1:58 , Lukas Odzioba <lukas.odzioba@...il.com> wrote:
> <quote>
> In the case of ”John the Ripper” [1] by which our implementation was
> inspired, sourcecode is available for review. On the Xeon machine used
> (see figure 1) John reaches about 7.3 million MD5 iterations per second
> in a single thread, whereas our 64 bit implementation reaches 33.4
> million iterations per second. However, there are some differences
> which might affect performance, because John the Ripper builds its
> passwords by a more complicated grammar using dictionaries.
> An earlier unpublished implementation by the authors of this paper
> using OpenSSL performed around 0.8 million iterations per second.
> </quote>
> source: http://wwwcip.cs.fau.de/~spjsschl/md5.pdf
> Source code is not available, we do not know what Xeon was tested, but
> Is there something that we could make use of?
That paper is early 2009... I think 7.3 million must have been 32bit non-SSE2, and not even x2? I get 38 million from current md5crypt using my 2.3 GHz laptop (37920 c/s).
magnum
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
