Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 14 Sep 2012 00:26:04 +0400
From: Solar Designer <>
Subject: Re: bitslice DES on GPU

On Wed, Sep 05, 2012 at 07:22:40AM +0400, Solar Designer wrote:
> To see how much of the slowness is from the extra copying and such
> rather than from the inner loop being slow, I tried increasing the
> iteration count from 25 to 725 and using this test hash:
> 	{"..X8NBuQ4l6uQ", ""},
> It's not a valid DES-based crypt(3) hash because those use 25 iterations
> (not configurable), but I generated it for my performance testing anyway.

I was wrong in that it was "not a valid DES-based crypt(3) hash".  It is.
Due to my use of an empty string for the password, the hash is valid for
any odd iteration count (e.g., 1, 25, 725, 2501, any other), because ""
corresponds to all-zero DES key, which is one of DES weak keys.  Thus,
every iteration of DES cancels the previous iteration, and only the
effect of the last iteration is left (the block is all-zero on input to
the last iteration).

This makes the hash above convenient for performance testing: we can
adjust the iteration count any way we like, and as long as we keep it
odd, the self-test will pass.


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.