Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Sat, 23 May 2015 03:41:19 +0300
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Agnieszka's weekly report #3

On Sat, May 23, 2015 at 02:00:50AM +0200, Agnieszka Bielec wrote:
> 2015-05-23 1:43 GMT+02:00 Agnieszka Bielec <bielecagnieszka8@...il.com>:
> > 2015-05-23 1:21 GMT+02:00 Solar Designer <solar@...nwall.com>:
> >> For Parallel, I hope you've noticed this? -
> >>
> >> http://www.openwall.com/lists/john-dev/2015/05/17/1
> >
> > I noticed
> >
> >> This is for CPU.  There's no need for you to duplicate Steve's work;
> >> rather, you need to port and integrate it.
> >
> > this is sha256 so this might require more work
> 
> I'm not even sure if I can use this

Oh, right.  Parallel submitted to PHC uses SHA-512, and this uses
SHA-256.  I think it makes sense for us to have both, much like we have
PBKDF2 with both SHA-256 and SHA-512, and more.  Parallel is in the same
category with PBKDF2 - a higher-level scheme.

BTW, you'll have better luck getting Parallel with SHA-256 run fast on
GPU (than for Parallel with SHA-512).

Steve's implementation of Parallel would bring in his own SHA-256 code,
though.  It's good code (includes interleaving and SIMD, all the way up
to AVX-512), but ideally we'd make our shared SHA-256 code at least as
good (possibly it already is, except for lacking use of AVX-512's
ternary logic intrinsics).  At least you can take a look at Steve's code
and learn from it.

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.