Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20150404150539.GA31708@openwall.com>
Date: Sat, 4 Apr 2015 18:05:39 +0300
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: [GSoC] John the Ripper support for PHC finalists

On Sat, Apr 04, 2015 at 04:52:47PM +0200, Frank Dittrich wrote:
> On 04/04/2015 04:44 PM, Agnieszka Bielec wrote:
> > thanks but what about --test command? before --test reset() is called
> > with db=NULL and tests are for fixed m_cost and t_cost
> 
> What exactly is your question here?
> 
> Unless you implement tunable cost reporting for pomelo and use --cost=,
> --test will check whether all the tests pass, and then use just the
> first two tests to compute the performance.
> 
> Pomelo isn't the only format that has tunable costs and self tests with
> varying values for the different tunable cost parameters.

Frank is right.

As to how to choose those first two test vectors, I suggest that we
choose either the lowest cost settings recommended by the author of
POMELO or cost settings that result in similar defensive throughput
capacity to bcrypt's cost 5 (in hashes per second when computing as many
concurrent hashes as the current CPUs support natively - so e.g. 8 on a
typical Intel quad-core with HT).  The latter will be convenient for us
to test POMELO's GPU attack resistance vs. bcrypt's.

Alexander

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.