Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <67cc89cad9225b2134a7ea3b183dd628@smtp.hushmail.com>
Date: Wed, 17 Oct 2012 20:01:53 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: NTLMv2 GPU

On 17 Apr, 2012, at 4:38 , jmk <jmk@...fus.net> wrote:
> On Mon, 2012-04-16 at 09:31 +0200, Lukas Odzioba wrote:
>> Hi all.
>> I've got an email from john user asking for NTLMv2 support for gpu's
>> (nVidia to be specific).
>> Are there any volounteers?
>> 
>> I could do this but if someone else loves this format I won't be an
>> obstacle :)
>> Lukas
> 
> I don't know if "love" is the right word, but I certainly use this
> format frequently. I'd be happy to test whatever code you create and
> answer any questions there may be about the code I originally hacked
> together for the format.

I have implemented this. It's noticeably I/O-bound on GPU, but on decent hardware the performance is awesome compared to the old CPU format, even with a weak GPU. And OpenCL/CPU outperforms OpenMP by far: The latter scales very poorly for this format and the new format runs vectorized on CPU while we have yet to support SSE2 in the old one.

It passes Test Suite (actually the TS found a subtle bug that could have gone under the radar forever) and has full Unicode/codepage support.

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.