Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 30 Jun 2011 03:50:33 +0200
From: Łukasz Odzioba <lukas.odzioba@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: Lukas's Status Report - #7 of 15

> Sounds good.  What's the status on md5crypt, though?
I've created 0-revision (simple raw, easy to understand, without
"dirty" optimizations)
It's already uploaded on wiki (few minutes ago), and benchmarks
results are on wiki gpu subpage.


> And can you please upload your "finished second version sha256crypt"
> right now (not next week)?  In general, I want code contributions to be
> available to the community sooner rather than later (which introduces
> extra risks - e.g., your hard drive might fail).

I'm doing code backups after grater changes, and often do not want to
share dug up code, hard to read, use for anyone besides me.
1-revision of sha256crypt was rewritten from scratch. It's much easier
to read and 25% faster than previous.

Results are highly depend on salt and pass length
For example:
saltlen=10,passlen=12
Old: 5578 c/s New: 6813 c/s

saltlen=10,passlen=4
Old: 10561 c/s New: l3248 c/s

I've added #FULL_UNROLL definition which unrolls completely main
sha256 loop and gives an additional 5% speedup (compilation takes 5
minutes longer).
I need to buy better power supply to test 9800gt, and add multigpu
support in the future, but while I'm unable to achieve much better
perfomance on single gpu it's not a high priority task.

Powered by blists - more mailing lists

Your e-mail address:

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