Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 17 Jun 2013 08:25:10 +0400
From: Solar Designer <>
Subject: Re: PBKDF2-SHA512

On Mon, Jun 17, 2013 at 06:16:42AM +0200, Lukas Odzioba wrote:
> 2013/6/17 Solar Designer <>:
> > I think you're comparing one core with C+intrinsics vs. all cores with
> > OpenCL.  Try building with OpenMP enabled for a more relevant comparison.
> I am not even using OpenCL. I attached the code just to show the idea
> - it does not have proper valid and other problems - to crappy and not
> heavily tested to be commited.
> Also I belive that speed could be improved by at least 10% using our
> pbkdf2-sha1 tricks and then doubled by proper sse code.

Oh, so you're saying that both of your benchmarks were for one core, and
that our pbkdf2_sha512_sse() in bleeding still does not implement all of
the usual PBKDF2 tricks, even though it does bother to use SIMD and
OpenMP.  Is that correct?  If so, we should optimize pbkdf2_sha512_sse()
further.  (I haven't looked at the code yet.)


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.