Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 02 Jul 2012 03:10:49 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: XSHA512 in Test Suite

On 2012-07-02 01:02, magnum wrote:
> On 2012-07-01 03:42, myrice wrote:
>> We don't have xsha512.in in TS. I also tested
>> them(xsha512-cuda/opencl) and believe them also reliable now. So can
>> we remove  "unreliable, will miss guesses" in these format also?
> 
> We do now. CUDA passes in 13 seconds and the "unreliable" is now removed
> from the format name. I'm still testing OpenCL. For some reason it takes
> AGES.
> 
> A problem is that TS loads 1500 passwords with 1500 different salts and
> the dictionary that will crack all of them fits in your GWS of 2M. This
> is not very realistic and is known to slow down some other test cases
> too, even on CPU. But why doesn't the CUDA format suffer much from this
> at all? It has a mkpc of 512K.

It actually takes like 3.5 hours vs CUDA's 26 seconds. Something is not
right.

form=xsha512-cuda                 guesses: 1468 time: 0:00:00:13  [PASSED]
.pot CHK:xsha512-cuda             guesses: 1468 time: 0:00:00:13  [PASSED]

form=xsha512-opencl               guesses: 1050 time: 0:01:47:10  [PASSED]

The .pot file check will likely take 1:47 also, so totalling 3:34.

However, I now removed the "unreliable" tag for that format too.

magnum

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ