Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Mon, 11 Jun 2012 20:07:26 +0400
From: Solar Designer <>
Cc: Per Thorsheim <>,
Subject: Re: MD5-crypt on Nvidia GPUs tips

On Sun, Apr 01, 2012 at 03:03:03PM +0400, Solar Designer wrote:
> Martijn - you show some weird speed number (not even a number) for John
> the Ripper.  I wonder where you got that one?  Well, my guess is that
> it's a non-OpenMP build of non-jumbo, or/and maybe a non-optimal make
> target.  We have plenty of faster speeds here:

I think I've figured it out.  Someone posted a link to the full
thesis.pdf to  That link is:

This is well-written overall, and it happens to give some details on the
hardware and software setup, as well as some dates in the references to
John the Ripper (2010, Jan 2011).  So if that's 2010 and Windows, I
understand why the low speed.  We only support 32-bit builds for
Windows (no 64-bit Windows support yet, although indeed we normally do
64-bit builds for Linux and many other systems), and there was in fact
no OpenMP support for MD5-crypt in 2010.

Yet the SHARCS 2012 presentation doesn't mention that it reuses speed
numbers from 2010, that those are for a 32-bit build, and that only one
CPU core was benchmarked.  In fact, IIRC, the paper submitted to SHARCS
2012 explicitly said that 4 cores were in use, yet gave a speed number
for 2010's 32-bit code on one core.

Whatever.  At least this appears clear now.

More importantly, thesis.pdf includes this link:

which is presumably where we could download the CUDA code and see how it
compares to Lukas' code currently in magnum-jumbo. resolves to an IP address in the Netherlands,
which unfortunately does not respond.

Martijn - would you share your code, please?  Preferably under a
suitable license, although indeed we don't yet know if we'd want to
reuse anything or not.



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.